AI/rocm-systems
2
0
Forka 0
Codice Problemi Pull Request Actions Pacchetti Progetti Rilasci Wiki Attività
Files
fa6f0717518c674995a3f09393214629c8565f59
rocm-systems/projects/amdsmi/amdsmi_cli/amdsmi_commands.py
T
Bindhiya Kanangot Balakrishnan fa6f071751 [SWDEV-574637] Avoid redundant hive gpu resets (#2657) 
Mode-1 GPU reset affects entire XGMI hive. Added
xgmi_hive_id check to reset only once for same-hive
GPUs while preserving separate resets for different
hives or no hives.
 - Example:
   `sudo amd-smi reset -G` or `sudo amd-smi reset -G -g 0`
   on MI300 will reset all GPU's only once.

Signed-off-by: Bindhiya Kanangot Balakrishnan <Bindhiya.KanangotBalakrishnan@amd.com>
2026-01-28 22:59:17 -06:00

7947 righe
435 KiB
Python
Originale Blame Cronologia

#!/usr/bin/env python3
#
# Copyright (C) Advanced Micro Devices. All rights reserved.
#
# Permission is hereby granted, free of charge, to any person obtaining a copy of
# this software and associated documentation files (the "Software"), to deal in
# the Software without restriction, including without limitation the rights to
# use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of
# the Software, and to permit persons to whom the Software is furnished to do so,
# subject to the following conditions:
#
# The above copyright notice and this permission notice shall be included in all
# copies or substantial portions of the Software.
#
# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
# IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS
# FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR
# COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER
# IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
# CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
import argparse
import json
import logging
import multiprocessing
import os
import signal
import sys
import threading
import time
import copy
from _version import __version__
from amdsmi_cli_exceptions import AmdSmiInvalidParameterException, AmdSmiRequiredCommandException, AmdSmiInvalidCommandException
from amdsmi_helpers import AMDSMIHelpers
from amdsmi_logger import AMDSMILogger
from amdsmi import amdsmi_exception, amdsmi_interface
from pathlib import Path
class AMDSMICommands():
"""This class contains all the commands corresponding to AMDSMIParser
Each command function will interact with AMDSMILogger to handle
displaying the output to the specified format and destination.
"""
def __init__(self, format='human_readable', destination='stdout', helpers=None) -> None:
if helpers is None:
# If helpers is not provided, create a new instance
self.helpers = AMDSMIHelpers()
else:
self.helpers = helpers
self.logger = AMDSMILogger(format=format, destination=destination, helpers=self.helpers)
self.device_handles = []
self.cpu_handles = []
self.core_handles = []
self.node_handle = None
self.stop = ''
self.group_check_printed = False
amdsmi_init_flag = self.helpers.get_amdsmi_init_flag()
logging.debug(f"AMDSMI Init Flag: {amdsmi_init_flag}")
exit_flag = False
if self.helpers.is_amdgpu_initialized():
try:
self.device_handles = amdsmi_interface.amdsmi_get_processor_handles()
except amdsmi_exception.AmdSmiLibraryException as e:
if e.err_code in (amdsmi_interface.amdsmi_wrapper.AMDSMI_STATUS_NOT_INIT,
amdsmi_interface.amdsmi_wrapper.AMDSMI_STATUS_DRIVER_NOT_LOADED):
logging.error('Unable to get devices, driver not initialized (amdgpu not found in modules)')
else:
raise e
if len(self.device_handles) == 0:
# No GPU's found post amdgpu driver initialization
logging.error('Unable to detect any GPU devices, check amdgpu version and module status (sudo modprobe amdgpu)')
exit_flag = True
# Resolve the node handle.
for dev in self.device_handles:
try:
nh = amdsmi_interface.amdsmi_get_node_handle(dev)
if nh is not None:
self.node_handle = nh
# Only need one handle, break after first success
break
except amdsmi_exception.AmdSmiLibraryException as e:
logging.debug("Unable to get node handle: %s", e.get_error_info())
# Node handle functionality is optional, so don't raise an error
if self.helpers.is_amd_hsmp_initialized():
try:
self.cpu_handles = amdsmi_interface.amdsmi_get_cpusocket_handles()
except amdsmi_exception.AmdSmiLibraryException as e:
if e.err_code in (amdsmi_interface.amdsmi_wrapper.AMDSMI_STATUS_NOT_INIT,
amdsmi_interface.amdsmi_wrapper.AMDSMI_STATUS_NO_DRV):
logging.info('Unable to detect any CPU devices, check amd_hsmp (or) hsmp_acpi version and module status (sudo modprobe amd_hsmp (or) sudo modprobe hsmp_acpi)')
else:
raise e
# core handles
try:
self.core_handles = amdsmi_interface.amdsmi_get_cpucore_handles()
except amdsmi_exception.AmdSmiLibraryException as e:
if e.err_code in (amdsmi_interface.amdsmi_wrapper.AMDSMI_STATUS_NOT_INIT,
amdsmi_interface.amdsmi_wrapper.AMDSMI_STATUS_NO_DRV):
logging.info('Unable to get CORE devices, amd_hsmp driver not loaded (sudo modprobe amd_hsmp)')
else:
raise e
if len(self.cpu_handles) == 0 and len(self.core_handles) == 0:
# No CPU's found post amd_hsmp driver initialization
logging.error('Unable to detect any CPU devices, check amd_hsmp (or) hsmp_acpi version and module status (sudo modprobe amd_hsmp (or) sudo modprobe hsmp_acpi)')
exit_flag = True
self.convert_clock_type = {
"sys": amdsmi_interface.AmdSmiClkType.SYS,
"mem": amdsmi_interface.AmdSmiClkType.MEM,
"df": amdsmi_interface.AmdSmiClkType.DF,
"soc": amdsmi_interface.AmdSmiClkType.SOC,
"dcef": amdsmi_interface.AmdSmiClkType.DCEF,
# vclk and dclk currently do not support levels so average clk is given for frequency levels
"vclk0": amdsmi_interface.AmdSmiClkType.VCLK0,
"vclk1": amdsmi_interface.AmdSmiClkType.VCLK1,
"dclk0": amdsmi_interface.AmdSmiClkType.DCLK0,
"dclk1": amdsmi_interface.AmdSmiClkType.DCLK1
}
if exit_flag:
version_args = argparse.Namespace()
version_args.gpu_version = False
version_args.cpu_version = False
self.version(version_args)
sys.exit(-1)
def version(self, args, gpu_version=None, cpu_version=None):
"""Print Version String
Args:
args (Namespace): Namespace containing the parsed CLI args
"""
if gpu_version:
args.gpu_version = gpu_version
if cpu_version:
args.cpu_version = cpu_version
# if no args are given, display everything
if args.gpu_version is None and args.cpu_version is None:
args.gpu_version = True
args.cpu_version = True
try:
amdsmi_lib_version = amdsmi_interface.amdsmi_get_lib_version()
amdsmi_lib_version_str = f"{amdsmi_lib_version['major']}.{amdsmi_lib_version['minor']}.{amdsmi_lib_version['release']}"
except amdsmi_exception.AmdSmiLibraryException as e:
amdsmi_lib_version_str = e.get_error_info()
try:
rocm_lib_status, rocm_version_str = amdsmi_interface.amdsmi_get_rocm_version()
if rocm_lib_status is not True:
rocm_version_str = "N/A"
except amdsmi_exception.AmdSmiLibraryException as e:
rocm_version_str = e.get_error_info()
self.logger.output['tool'] = 'AMDSMI Tool'
self.logger.output['version'] = f'{__version__}'
self.logger.output['amdsmi_library_version'] = f'{amdsmi_lib_version_str}'
self.logger.output['rocm_version'] = f'{rocm_version_str}'
if args.gpu_version:
try:
gpus = amdsmi_interface.amdsmi_get_processor_handles()
if isinstance(gpus, list) and len(gpus) > 0:
gpu_version_info = amdsmi_interface.amdsmi_get_gpu_driver_info(gpus[0])
gpu_version_str = gpu_version_info['driver_version']
else:
gpu_version_str = "N/A"
except amdsmi_exception.AmdSmiLibraryException as e:
gpu_version_str = e.get_error_info()
self.logger.output['amdgpu_version'] = gpu_version_str
if args.cpu_version:
try:
cpus = amdsmi_interface.amdsmi_get_cpusocket_handles()
if isinstance(cpus, list) and len(cpus) > 0:
cpu_version_info = amdsmi_interface.amdsmi_get_cpu_hsmp_driver_version(cpus[0])
cpu_version_str = str(cpu_version_info['hsmp_driver_major_ver_num']) + "." + str(cpu_version_info['hsmp_driver_minor_ver_num'])
else:
cpu_version_str = "N/A"
except amdsmi_exception.AmdSmiLibraryException as e:
cpu_version_str = e.get_error_info()
self.logger.output['amd_hsmp_driver_version'] = cpu_version_str
if self.logger.is_human_readable_format():
human_readable_output = f"AMDSMI Tool: {__version__} | " \
f"AMDSMI Library version: {amdsmi_lib_version_str} | " \
f"ROCm version: {rocm_version_str}"
if args.gpu_version:
human_readable_output = human_readable_output + f" | amdgpu version: {gpu_version_str}"
if args.cpu_version:
human_readable_output = human_readable_output + f" | hsmp version: {cpu_version_str}"
# Custom human readable handling for version
if self.logger.destination == 'stdout':
print(human_readable_output)
else:
with self.logger.destination.open('a', encoding="utf-8") as output_file:
output_file.write(human_readable_output + '\n')
elif self.logger.is_json_format() or self.logger.is_csv_format():
self.logger.print_output()
def list(self, args, multiple_devices=False, gpu=None):
"""List information for target gpu
Args:
args (Namespace): Namespace containing the parsed CLI args
multiple_devices (bool, optional): True if checking for multiple devices. Defaults to False.
gpu (device_handle, optional): device_handle for target device. Defaults to None.
Raises:
IndexError: Index error if gpu list is empty
Returns:
None: Print output via AMDSMILogger to destination
"""
# Set args.* to passed in arguments
if gpu:
args.gpu = gpu
cpu_attributes = ["cpu"]
for attr in cpu_attributes:
if hasattr(args, 'cpu') and getattr(args, 'cpu'):
print("N/A")
return
# Handle No GPU passed
if args.gpu == None:
args.gpu = self.device_handles
if not self.group_check_printed:
self.helpers.check_required_groups(check_render=True, check_video=False)
self.group_check_printed = True
# Handle multiple GPUs
handled_multiple_gpus, device_handle = self.helpers.handle_gpus(args, self.logger, self.list)
if handled_multiple_gpus:
return # This function is recursive
args.gpu = device_handle
# Get gpu_id for logging
gpu_id = self.helpers.get_gpu_id_from_device_handle(args.gpu)
# Always try to get BDF regardless of group check
try:
bdf = amdsmi_interface.amdsmi_get_gpu_device_bdf(args.gpu)
except amdsmi_exception.AmdSmiLibraryException as e:
bdf = "N/A"
try:
uuid = amdsmi_interface.amdsmi_get_gpu_device_uuid(args.gpu)
except amdsmi_exception.AmdSmiLibraryException:
uuid = "N/A"
try:
kfd_info = amdsmi_interface.amdsmi_get_gpu_kfd_info(args.gpu)
kfd_id = kfd_info['kfd_id']
node_id = kfd_info['node_id']
partition_id = kfd_info['current_partition_id']
except amdsmi_exception.AmdSmiLibraryException as e:
kfd_id = node_id = partition_id = "N/A"
logging.debug("Failed to get kfd info for gpu %s | %s", gpu_id, e.get_error_info())
# CSV format is intentionally aligned with Host
if self.logger.is_csv_format():
self.logger.store_output(args.gpu, 'gpu_bdf', bdf)
self.logger.store_output(args.gpu, 'gpu_uuid', uuid)
else:
self.logger.store_output(args.gpu, 'bdf', bdf)
self.logger.store_output(args.gpu, 'uuid', uuid)
self.logger.store_output(args.gpu, 'kfd_id', kfd_id)
self.logger.store_output(args.gpu, 'node_id', node_id)
self.logger.store_output(args.gpu, 'partition_id', partition_id)
if args.e:
try:
enumeration_info = amdsmi_interface.amdsmi_get_gpu_enumeration_info(args.gpu)
except amdsmi_exception.AmdSmiLibraryException:
enumeration_info = {
"drm_render": "N/A",
"drm_card": "N/A",
"hsa_id": "N/A",
"hip_id": "N/A",
"hip_uuid": "N/A",
}
# now store all the fields exactly once:
if enumeration_info['drm_render'] == "N/A":
self.logger.store_output(args.gpu, 'render', enumeration_info['drm_render'])
else:
self.logger.store_output(args.gpu, 'render',
f"renderD{enumeration_info['drm_render']}")
if enumeration_info['drm_card'] == "N/A":
self.logger.store_output(args.gpu, 'card', enumeration_info['drm_card'])
else:
self.logger.store_output(args.gpu, 'card',
f"card{enumeration_info['drm_card']}")
self.logger.store_output(args.gpu, 'hsa_id', enumeration_info['hsa_id'])
self.logger.store_output(args.gpu, 'hip_id', enumeration_info['hip_id'])
self.logger.store_output(args.gpu, 'hip_uuid', enumeration_info['hip_uuid'])
if multiple_devices:
self.logger.store_multiple_device_output()
return # Skip printing when there are multiple devices
self.logger.print_output()
def static_cpu(self, args, multiple_devices=False, cpu=None, interface_ver=None):
"""Get Static information for target cpu
Args:
args (Namespace): Namespace containing the parsed CLI args
multiple_devices (bool, optional): True if checking for multiple devices. Defaults to False.
cpu (device_handle, optional): device_handle for target device. Defaults to None.
Returns:
None: Print output via AMDSMILogger to destination
"""
if cpu:
args.cpu = cpu
if interface_ver:
args.interface_ver = interface_ver
# Store cpu args that are applicable to the current platform
curr_platform_cpu_args = ["smu", "interface_ver"]
curr_platform_cpu_values = [args.smu, args.interface_ver]
# If no cpu options are passed, return all available args
if not any(curr_platform_cpu_values):
for arg in curr_platform_cpu_args:
setattr(args, arg, True)
# Handle multiple CPUs
handled_multiple_cpus, device_handle = self.helpers.handle_cpus(args,
self.logger,
self.static_cpu)
if handled_multiple_cpus:
return # This function is recursive
args.cpu = device_handle
# Get cpu id for logging
cpu_id = self.helpers.get_cpu_id_from_device_handle(args.cpu)
logging.debug(f"Static Arg information for CPU {cpu_id} on {self.helpers.os_info()}")
static_dict = {}
if self.logger.is_json_format():
static_dict['cpu'] = int(cpu_id)
if args.smu:
try:
smu = amdsmi_interface.amdsmi_get_cpu_smu_fw_version(args.cpu)
static_dict["smu"] = {"FW_VERSION" : f"{smu['smu_fw_major_ver_num']}."
f"{smu['smu_fw_minor_ver_num']}.{smu['smu_fw_debug_ver_num']}"}
except amdsmi_exception.AmdSmiLibraryException as e:
static_dict["smu"] = "N/A"
logging.debug("Failed to get SMU FW for cpu %s | %s", cpu_id, e.get_error_info())
if args.interface_ver:
static_dict["interface_version"] = {}
try:
intf_ver = amdsmi_interface.amdsmi_get_cpu_hsmp_proto_ver(args.cpu)
static_dict["interface_version"]["proto version"] = intf_ver
except amdsmi_exception.AmdSmiLibraryException as e:
static_dict["interface_version"]["proto version"] = "N/A"
logging.debug("Failed to get proto version for cpu %s | %s", cpu_id, e.get_error_info())
multiple_devices_csv_override = False
if not self.logger.is_json_format():
self.logger.store_cpu_output(args.cpu, 'values', static_dict)
else:
self.logger.store_cpu_json_output.append(static_dict)
if multiple_devices:
self.logger.store_multiple_device_output()
return # Skip printing when there are multiple devices
if not self.logger.is_json_format():
self.logger.print_output(multiple_device_enabled=multiple_devices_csv_override)
def static_gpu(self, args, multiple_devices=False, gpu=None, asic=None, bus=None, vbios=None,
limit=None, driver=None, ras=None, board=None, numa=None, vram=None,
cache=None, partition=None, dfc_ucode=None, fb_info=None, num_vf=None,
soc_pstate=None, xgmi_plpd=None, process_isolation=None, clock=None, profile=None):
"""Get Static information for target gpu
Args:
args (Namespace): Namespace containing the parsed CLI args
current_platform_args (list): gpu supported platform arguments
current_platform_values (list): gpu supported platform values for each argument
multiple_devices (bool, optional): True if checking for multiple devices. Defaults to False.
gpu (device_handle, optional): device_handle for target device. Defaults to None.
asic (bool, optional): Value override for args.asic. Defaults to None.
bus (bool, optional): Value override for args.bus. Defaults to None.
vbios (bool, optional): Value override for args.vbios. Defaults to None.
limit (bool, optional): Value override for args.limit. Defaults to None.
driver (bool, optional): Value override for args.driver. Defaults to None.
ras (bool, optional): Value override for args.ras. Defaults to None.
board (bool, optional): Value override for args.board. Defaults to None.
numa (bool, optional): Value override for args.numa. Defaults to None.
vram (bool, optional): Value override for args.vram. Defaults to None.
cache (bool, optional): Value override for args.cache. Defaults to None.
partition (bool, optional): Value override for args.partition. Defaults to None.
dfc_ucode (bool, optional): Value override for args.dfc_ucode. Defaults to None.
fb_info (bool, optional): Value override for args.fb_info. Defaults to None.
num_vf (bool, optional): Value override for args.num_vf. Defaults to None.
soc_pstate (bool, optional): Value override for args.soc_pstate. Defaults to None.
xgmi_plpd (bool, optional): Value override for args.xgmi_plpd. Defaults to None.
process_isolation (bool, optional): Value override for args.process_isolation. Defaults to None.
Returns:
None: Print output via AMDSMILogger to destination
"""
if gpu:
args.gpu = gpu
if asic:
args.asic = asic
if bus:
args.bus = bus
if vbios:
args.vbios = vbios
if board:
args.board = board
if driver:
args.driver = driver
if ras:
args.ras = ras
if vram:
args.vram = vram
if cache:
args.cache = cache
if process_isolation:
args.process_isolation = process_isolation
if partition:
args.partition = partition
if clock:
args.clock = clock
# args.clock defaults to False so if it was overwritten to empty list, that indicates that it was given as an arguments but with an empty list
if args.clock == []:
args.clock = True
# Store args that are applicable to the current platform (default arguments)
current_platform_args = ["asic", "bus", "vbios", "driver", "ras",
"vram", "cache", "board", "process_isolation",
"clock"]
current_platform_values = [args.asic, args.bus, args.vbios, args.driver, args.ras,
args.vram, args.cache, args.board, args.process_isolation,
args.clock]
# amd-smi static default arguments:
# Exclude args that are not applicable to the current platform,
# but allow output if argument is passed.
#
# Note: Partition is a special case, it is no longer an amd-smi static
# default argument.
# Reason: Reading current_compute_partition may momentarily wake the
# GPU up. This is due to reading XCD registers, which is expected
# behavior. Changing partitions is not a trivial operation,
# current_compute_partition SYSFS controls this action.
if args.partition:
current_platform_args += ["partition"]
current_platform_values += [args.partition]
if not self.group_check_printed:
self.helpers.check_required_groups(check_render=True, check_video=False)
self.group_check_printed = True
if self.helpers.is_linux() and self.helpers.is_baremetal():
if limit:
args.limit = limit
if soc_pstate:
args.soc_pstate = soc_pstate
if xgmi_plpd:
args.xgmi_plpd = xgmi_plpd
if profile:
args.profile = profile
current_platform_args += ["ras", "limit", "soc_pstate", "xgmi_plpd", "profile"]
current_platform_values += [args.ras, args.limit, args.soc_pstate, args.xgmi_plpd, args.profile]
if self.helpers.is_linux() and not self.helpers.is_virtual_os():
if numa:
args.numa = numa
current_platform_args += ["numa"]
current_platform_values += [args.numa]
if self.helpers.is_hypervisor():
if dfc_ucode:
args.dfc_ucode = dfc_ucode
if fb_info:
args.fb_info = fb_info
if num_vf:
args.num_vf = num_vf
current_platform_args += ["dfc_ucode", "fb_info", "num_vf"]
current_platform_values += [args.dfc_ucode, args.fb_info, args.num_vf]
if not any(current_platform_values):
for arg in current_platform_args:
setattr(args, arg, True)
handled_multiple_gpus, device_handle = self.helpers.handle_gpus(args, self.logger, self.static_gpu)
if handled_multiple_gpus:
return # This function is recursive
args.gpu = device_handle
# Get gpu_id for logging
gpu_id = self.helpers.get_gpu_id_from_device_handle(args.gpu)
logging.debug("=====================================================================")
logging.debug(f"Static Arg information for GPU {gpu_id} on {self.helpers.os_info()}")
logging.debug(f"Function args: {args}")
logging.debug(f"Current platform args: {current_platform_args}")
logging.debug(f"Current platform values: {current_platform_values}")
logging.debug("=====================================================================")
# Populate static dictionary for each enabled argument
static_dict = {}
if self.logger.is_json_format():
static_dict['gpu'] = int(gpu_id)
if args.asic:
asic_dict = {
"market_name" : "N/A",
"vendor_id" : "N/A",
"vendor_name" : "N/A",
"subvendor_id" : "N/A",
"device_id" : "N/A",
"subsystem_id" : "N/A",
"rev_id" : "N/A",
"asic_serial" : "N/A",
"oam_id" : "N/A",
"num_compute_units" : "N/A",
"target_graphics_version" : "N/A"
}
try:
asic_info = amdsmi_interface.amdsmi_get_gpu_asic_info(args.gpu)
for key, value in asic_info.items():
asic_dict[key] = value
except amdsmi_exception.AmdSmiLibraryException as e:
logging.debug("Failed to get asic info for gpu %s | %s", gpu_id, e.get_error_info())
static_dict['asic'] = asic_dict
if args.bus:
bus_info = {
'bdf': "N/A",
'max_pcie_width': "N/A",
'max_pcie_speed': "N/A",
'pcie_levels': "N/A",
'pcie_interface_version': "N/A",
'slot_type': "N/A"
}
try:
bus_info['bdf'] = amdsmi_interface.amdsmi_get_gpu_device_bdf(args.gpu)
except amdsmi_exception.AmdSmiLibraryException as e:
bus_info['bdf'] = "N/A"
logging.debug("Failed to get bdf for gpu %s | %s", gpu_id, e.get_error_info())
try:
pcie_static = amdsmi_interface.amdsmi_get_pcie_info(args.gpu)['pcie_static']
bus_info['max_pcie_width'] = pcie_static['max_pcie_width']
bus_info['max_pcie_speed'] = pcie_static['max_pcie_speed']
bus_info['pcie_interface_version'] = pcie_static['pcie_interface_version']
bus_info['slot_type'] = pcie_static['slot_type']
if bus_info['max_pcie_speed'] % 1000 != 0:
pcie_speed_GTs_value = round(bus_info['max_pcie_speed'] / 1000, 1)
else:
pcie_speed_GTs_value = round(bus_info['max_pcie_speed'] / 1000)
bus_info['max_pcie_speed'] = pcie_speed_GTs_value
if bus_info['pcie_interface_version'] > 0:
bus_info['pcie_interface_version'] = f"Gen {bus_info['pcie_interface_version']}"
# Set the unit for pcie_speed
pcie_speed_unit ='GT/s'
if self.logger.is_human_readable_format():
bus_info['max_pcie_speed'] = f"{bus_info['max_pcie_speed']} {pcie_speed_unit}"
if self.logger.is_json_format():
bus_info['max_pcie_speed'] = {"value" : bus_info['max_pcie_speed'],
"unit" : pcie_speed_unit}
except amdsmi_exception.AmdSmiLibraryException as e:
logging.debug("Failed to get bus info for gpu %s | %s", gpu_id, e.get_error_info())
try:
pcie_info = amdsmi_interface.amdsmi_get_gpu_pci_bandwidth(args.gpu)
num_supported = pcie_info['transfer_rate']['num_supported']
if num_supported != 0:
bus_info['pcie_levels'] = {}
for level in range(0, num_supported):
speed = str(self.helpers.convert_SI_unit(float(pcie_info['transfer_rate']['frequency'][level]), AMDSMIHelpers.SI_Unit.NANO)) + " GT/s"
width = str(pcie_info['lanes'][level])
level_values = (speed, width)
bus_info['pcie_levels'].update({str(level): level_values})
else:
bus_info['pcie_levels'] = "N/A"
except amdsmi_exception.AmdSmiLibraryException as e:
logging.debug("Failed to get pci bandwidth info for gpu %s | %s", gpu_id, e.get_error_info())
static_dict['bus'] = bus_info
if args.vbios:
try:
vbios_info = amdsmi_interface.amdsmi_get_gpu_vbios_info(args.gpu)
for key, value in vbios_info.items():
if isinstance(value, str):
if value.strip() == '':
vbios_info[key] = "N/A"
static_dict['ifwi'] = vbios_info
# Remove boot_firmware since it's not used
del static_dict['ifwi']['boot_firmware']
except amdsmi_exception.AmdSmiLibraryException as e:
static_dict['ifwi'] = "N/A"
logging.debug("Failed to get vbios/ifwi info for gpu %s | %s", gpu_id, e.get_error_info())
if 'limit' in current_platform_args:
if args.limit:
# Power limits
power_limit_types = {}
for power_type in amdsmi_interface.AmdSmiPowerCapType:
# Strip 'AMDSMI_POWER_CAP_TYPE_' prefix and convert to lowercase
key = power_type.name.replace('AMDSMI_POWER_CAP_TYPE_', '').lower()
power_limit_types[key] = {
"max_power_limit" : "N/A",
"min_power_limit" : "N/A",
"socket_power_limit" : "N/A"
}
try:
power_cap_types = amdsmi_interface.amdsmi_get_supported_power_cap(args.gpu)
for sensor in power_cap_types['sensor_inds']:
power_cap_info = amdsmi_interface.amdsmi_get_power_cap_info(args.gpu, sensor)
max_power_limit = power_cap_info['max_power_cap']
max_power_limit = self.helpers.convert_SI_unit(max_power_limit, AMDSMIHelpers.SI_Unit.MICRO)
min_power_limit = power_cap_info['min_power_cap']
min_power_limit = self.helpers.convert_SI_unit(min_power_limit, AMDSMIHelpers.SI_Unit.MICRO)
socket_power_limit = power_cap_info['power_cap']
socket_power_limit = self.helpers.convert_SI_unit(socket_power_limit, AMDSMIHelpers.SI_Unit.MICRO)
ppt = {
"max_power_limit" : self.helpers.unit_format(self.logger, max_power_limit, 'W'),
"min_power_limit" : self.helpers.unit_format(self.logger, min_power_limit, 'W'),
"socket_power_limit" : self.helpers.unit_format(self.logger, socket_power_limit, 'W')
}
sensor_name = power_cap_types['sensor_types'][sensor]
# Strip 'AMDSMI_POWER_CAP_TYPE_' prefix and convert to lowercase
sensor_key = sensor_name.name.replace('AMDSMI_POWER_CAP_TYPE_', '').lower()
power_limit_types[sensor_key] = ppt
except amdsmi_exception.AmdSmiLibraryException as e:
logging.debug("Failed to get power cap info for gpu %s | %s", gpu_id, e.get_error_info())
# Edge temperature limits
try:
slowdown_temp_edge_limit_error = False
slowdown_temp_edge_limit = amdsmi_interface.amdsmi_get_temp_metric(args.gpu,
amdsmi_interface.AmdSmiTemperatureType.EDGE, amdsmi_interface.AmdSmiTemperatureMetric.CRITICAL)
except amdsmi_exception.AmdSmiLibraryException as e:
slowdown_temp_edge_limit_error = True
slowdown_temp_edge_limit = "N/A"
logging.debug("Failed to get edge temperature slowdown metric for gpu %s | %s", gpu_id, e.get_error_info())
if slowdown_temp_edge_limit == 0:
slowdown_temp_edge_limit_error = True
slowdown_temp_edge_limit = "N/A"
try:
shutdown_temp_edge_limit_error = False
shutdown_temp_edge_limit = amdsmi_interface.amdsmi_get_temp_metric(args.gpu,
amdsmi_interface.AmdSmiTemperatureType.EDGE, amdsmi_interface.AmdSmiTemperatureMetric.EMERGENCY)
except amdsmi_exception.AmdSmiLibraryException as e:
shutdown_temp_edge_limit_error = True
shutdown_temp_edge_limit = "N/A"
logging.debug("Failed to get edge temperature shutdown metrics for gpu %s | %s", gpu_id, e.get_error_info())
if shutdown_temp_edge_limit == 0:
shutdown_temp_edge_limit_error = True
shutdown_temp_edge_limit = "N/A"
# Hotspot/Junction temperature limits
try:
slowdown_temp_hotspot_limit_error = False
slowdown_temp_hotspot_limit = amdsmi_interface.amdsmi_get_temp_metric(args.gpu,
amdsmi_interface.AmdSmiTemperatureType.HOTSPOT, amdsmi_interface.AmdSmiTemperatureMetric.CRITICAL)
except amdsmi_exception.AmdSmiLibraryException as e:
slowdown_temp_hotspot_limit_error = True
slowdown_temp_hotspot_limit = "N/A"
logging.debug("Failed to get hotspot temperature slowdown metrics for gpu %s | %s", gpu_id, e.get_error_info())
try:
shutdown_temp_hotspot_limit_error = False
shutdown_temp_hotspot_limit = amdsmi_interface.amdsmi_get_temp_metric(args.gpu,
amdsmi_interface.AmdSmiTemperatureType.HOTSPOT, amdsmi_interface.AmdSmiTemperatureMetric.EMERGENCY)
except amdsmi_exception.AmdSmiLibraryException as e:
shutdown_temp_hotspot_limit_error = True
shutdown_temp_hotspot_limit = "N/A"
logging.debug("Failed to get hotspot temperature shutdown metrics for gpu %s | %s", gpu_id, e.get_error_info())
# VRAM temperature limits
try:
slowdown_temp_vram_limit_error = False
slowdown_temp_vram_limit = amdsmi_interface.amdsmi_get_temp_metric(args.gpu,
amdsmi_interface.AmdSmiTemperatureType.VRAM, amdsmi_interface.AmdSmiTemperatureMetric.CRITICAL)
except amdsmi_exception.AmdSmiLibraryException as e:
slowdown_temp_vram_limit_error = True
slowdown_temp_vram_limit = "N/A"
logging.debug("Failed to get vram temperature slowdown metrics for gpu %s | %s", gpu_id, e.get_error_info())
try:
shutdown_temp_vram_limit_error = False
shutdown_temp_vram_limit = amdsmi_interface.amdsmi_get_temp_metric(args.gpu,
amdsmi_interface.AmdSmiTemperatureType.VRAM, amdsmi_interface.AmdSmiTemperatureMetric.EMERGENCY)
except amdsmi_exception.AmdSmiLibraryException as e:
shutdown_temp_vram_limit_error = True
shutdown_temp_vram_limit = "N/A"
logging.debug("Failed to get vram temperature shutdown metrics for gpu %s | %s", gpu_id, e.get_error_info())
# PTL
try:
ptl_state = amdsmi_interface.amdsmi_get_gpu_ptl_state(args.gpu)
ptl_state = "Enabled" if ptl_state else "Disabled"
except amdsmi_exception.AmdSmiLibraryException as e:
ptl_state = "N/A"
logging.debug("Failed to get PTL state for gpu %s | %s", gpu_id, e.get_error_info())
try:
ptl_format1, ptl_format2 = amdsmi_interface.amdsmi_get_gpu_ptl_formats(args.gpu)
fmt1_name = amdsmi_interface.amdsmi_wrapper.amdsmi_ptl_data_format_t__enumvalues.get(ptl_format1)
fmt2_name = amdsmi_interface.amdsmi_wrapper.amdsmi_ptl_data_format_t__enumvalues.get(ptl_format2)
fmt1_short = fmt1_name.replace("AMDSMI_PTL_DATA_FORMAT_", "") if fmt1_name else "UNKNOWN"
fmt2_short = fmt2_name.replace("AMDSMI_PTL_DATA_FORMAT_", "") if fmt2_name else "UNKNOWN"
ptl_format = f"{fmt1_short},{fmt2_short}"
except amdsmi_exception.AmdSmiLibraryException as e:
ptl_format = "N/A"
logging.debug("Failed to get PTL state for gpu %s | %s", gpu_id, e.get_error_info())
# Assign units
power_unit = 'W'
temp_unit_human_readable = '\N{DEGREE SIGN}C'
temp_unit_json = 'C'
if self.logger.is_human_readable_format():
if not slowdown_temp_edge_limit_error:
slowdown_temp_edge_limit = f"{slowdown_temp_edge_limit} {temp_unit_human_readable}"
if not slowdown_temp_hotspot_limit_error:
slowdown_temp_hotspot_limit = f"{slowdown_temp_hotspot_limit} {temp_unit_human_readable}"
if not slowdown_temp_vram_limit_error:
slowdown_temp_vram_limit = f"{slowdown_temp_vram_limit} {temp_unit_human_readable}"
if not shutdown_temp_edge_limit_error:
shutdown_temp_edge_limit = f"{shutdown_temp_edge_limit} {temp_unit_human_readable}"
if not shutdown_temp_hotspot_limit_error:
shutdown_temp_hotspot_limit = f"{shutdown_temp_hotspot_limit} {temp_unit_human_readable}"
if not shutdown_temp_vram_limit_error:
shutdown_temp_vram_limit = f"{shutdown_temp_vram_limit} {temp_unit_human_readable}"
if self.logger.is_json_format():
if not slowdown_temp_edge_limit_error:
slowdown_temp_edge_limit = {"value" : slowdown_temp_edge_limit,
"unit" : temp_unit_json}
if not slowdown_temp_hotspot_limit_error:
slowdown_temp_hotspot_limit = {"value" : slowdown_temp_hotspot_limit,
"unit" : temp_unit_json}
if not slowdown_temp_vram_limit_error:
slowdown_temp_vram_limit = {"value" : slowdown_temp_vram_limit,
"unit" : temp_unit_json}
if not shutdown_temp_edge_limit_error:
shutdown_temp_edge_limit = {"value" : shutdown_temp_edge_limit,
"unit" : temp_unit_json}
if not shutdown_temp_hotspot_limit_error:
shutdown_temp_hotspot_limit = {"value" : shutdown_temp_hotspot_limit,
"unit" : temp_unit_json}
if not shutdown_temp_vram_limit_error:
shutdown_temp_vram_limit = {"value" : shutdown_temp_vram_limit,
"unit" : temp_unit_json}
limit_info = {}
# Power limits
limit_info['ppt0'] = power_limit_types['ppt0']
limit_info['ppt1'] = power_limit_types['ppt1']
# Shutdown limits
limit_info['slowdown_edge_temperature'] = slowdown_temp_edge_limit
limit_info['slowdown_hotspot_temperature'] = slowdown_temp_hotspot_limit
limit_info['slowdown_vram_temperature'] = slowdown_temp_vram_limit
limit_info['shutdown_edge_temperature'] = shutdown_temp_edge_limit
limit_info['shutdown_hotspot_temperature'] = shutdown_temp_hotspot_limit
limit_info['shutdown_vram_temperature'] = shutdown_temp_vram_limit
# PTL
limit_info['ptl_state'] = ptl_state
limit_info['ptl_format'] = ptl_format
static_dict['limit'] = limit_info
if args.driver:
driver_info_dict = {"name" : "N/A",
"version" : "N/A",
"os_kernel_version" : "N/A"}
try:
driver_info = amdsmi_interface.amdsmi_get_gpu_driver_info(args.gpu)
driver_info_dict["name"] = driver_info["driver_name"]
driver_info_dict["version"] = driver_info["driver_version"]
except amdsmi_exception.AmdSmiLibraryException as e:
logging.debug("Failed to get driver info for gpu %s | %s", gpu_id, e.get_error_info())
try:
driver_info_dict["os_kernel_version"] = os.uname().release
except (AttributeError, OSError) as e:
logging.debug("Failed to get os kernel version for gpu %s | %s", gpu_id, e)
static_dict['driver'] = driver_info_dict
if args.board:
static_dict['board'] = {"model_number": "N/A",
"product_serial": "N/A",
"fru_id": "N/A",
"product_name": "N/A",
"manufacturer_name": "N/A"}
try:
board_info = amdsmi_interface.amdsmi_get_gpu_board_info(args.gpu)
for key, value in board_info.items():
if isinstance(value, str):
if value.strip() == '':
board_info[key] = "N/A"
static_dict['board'] = board_info
except amdsmi_exception.AmdSmiLibraryException as e:
logging.debug("Failed to get board info for gpu %s | %s", gpu_id, e.get_error_info())
if 'ras' in current_platform_args:
if args.ras:
ras_dict = {"eeprom_version": "N/A",
"bad_page_threshold": "N/A",
"bad_page_threshold_exceeded": "N/A",
"parity_schema" : "N/A",
"single_bit_schema" : "N/A",
"double_bit_schema" : "N/A",
"poison_schema" : "N/A",
"ecc_block_state": "N/A"}
try:
ras_info = amdsmi_interface.amdsmi_get_gpu_ras_feature_info(args.gpu)
for key, value in ras_info.items():
if isinstance(value, int):
if value == 65535:
logging.debug(f"Failed to get ras {key} for gpu {gpu_id}")
ras_info[key] = "N/A"
continue
if key != "eeprom_version":
if value:
ras_info[key] = "ENABLED"
else:
ras_info[key] = "DISABLED"
ras_dict.update(ras_info)
except amdsmi_exception.AmdSmiLibraryException as e:
logging.debug("Failed to get ras info for gpu %s | %s", gpu_id, e.get_error_info())
try:
ras_dict["bad_page_threshold"] = amdsmi_interface.amdsmi_get_gpu_bad_page_threshold(args.gpu)
except amdsmi_exception.AmdSmiLibraryException as e:
logging.debug("Failed to get bad page threshold count for gpu %s | %s", gpu_id, e.get_error_info())
try:
bad_page_info = amdsmi_interface.amdsmi_get_gpu_bad_page_info(args.gpu)
retired_pages = 0
if bad_page_info:
for bad_page in bad_page_info:
if bad_page["status"] == amdsmi_interface.AmdSmiMemoryPageStatus.RESERVED:
retired_pages += 1
# default to N/A
ras_dict["bad_page_threshold_exceeded"] = "N/A"
# If this is an int, then default to False
if isinstance(ras_dict["bad_page_threshold"], int):
ras_dict["bad_page_threshold_exceeded"] = "False"
if retired_pages > ras_dict["bad_page_threshold"]:
# If there are more retired pages then set to True
ras_dict["bad_page_threshold_exceeded"] = "True"
except amdsmi_exception.AmdSmiLibraryException as e:
logging.debug("Failed to get retired pages count for gpu %s | %s", gpu_id, e.get_error_info())
try:
ras_states = amdsmi_interface.amdsmi_get_gpu_ras_block_features_enabled(args.gpu)
ecc_block_state_dict = {}
for state in ras_states:
ecc_block_state_dict[state["block"]] = state["status"]
ras_dict["ecc_block_state"] = ecc_block_state_dict
except amdsmi_exception.AmdSmiLibraryException as e:
logging.debug("Failed to get ras block features for gpu %s | %s", gpu_id, e.get_error_info())
static_dict["ras"] = ras_dict
if args.partition:
try:
compute_partition = amdsmi_interface.amdsmi_get_gpu_compute_partition(args.gpu)
except amdsmi_exception.AmdSmiLibraryException as e:
compute_partition = "N/A"
logging.debug("Failed to get compute partition info for gpu %s | %s", gpu_id, e.get_error_info())
try:
memory_partition = amdsmi_interface.amdsmi_get_gpu_memory_partition(args.gpu)
except amdsmi_exception.AmdSmiLibraryException as e:
memory_partition = "N/A"
logging.debug("Failed to get memory partition info for gpu %s | %s", gpu_id, e.get_error_info())
try:
kfd_info = amdsmi_interface.amdsmi_get_gpu_kfd_info(args.gpu)
partition_id = kfd_info['current_partition_id']
except amdsmi_exception.AmdSmiLibraryException as e:
partition_id = "N/A"
logging.debug("Failed to get partition ID for gpu %s | %s", gpu_id, e.get_error_info())
static_dict['partition'] = {"accelerator_partition": compute_partition,
"memory_partition": memory_partition,
"partition_id": partition_id}
if 'soc_pstate' in current_platform_args:
if args.soc_pstate:
try:
policy_info = amdsmi_interface.amdsmi_get_soc_pstate(args.gpu)
except amdsmi_exception.AmdSmiLibraryException as e:
policy_info = "N/A"
logging.debug("Failed to get soc pstate policy info for gpu %s | %s", gpu_id, e.get_error_info())
# Format for CSV output - flatten completely to avoid extra columns
if self.logger.is_csv_format() and isinstance(policy_info, dict):
policies_str = ', '.join(
f"{p['policy_id']}:{p['policy_description']}"
for p in policy_info.get('policies', [])
) or 'N/A'
static_dict['num_supported'] = policy_info.get('num_supported', 'N/A')
static_dict['current_id'] = policy_info.get('current_id', 'N/A')
static_dict['policies'] = policies_str
else:
static_dict['soc_pstate'] = policy_info
if 'xgmi_plpd' in current_platform_args:
if args.xgmi_plpd:
try:
policy_info = amdsmi_interface.amdsmi_get_xgmi_plpd(args.gpu)
except amdsmi_exception.AmdSmiLibraryException as e:
policy_info = "N/A"
logging.debug("Failed to get xgmi_plpd info for gpu %s | %s", gpu_id, e.get_error_info())
# Format for CSV output - flatten completely to avoid extra columns
if self.logger.is_csv_format() and isinstance(policy_info, dict):
policies_str = ', '.join(
f"{p['policy_id']}:{p['policy_description']}"
for p in policy_info.get('policies', [])
) or 'N/A'
static_dict['num_supported'] = policy_info.get('num_supported', 'N/A')
static_dict['current_id'] = policy_info.get('current_id', 'N/A')
static_dict['policies'] = policies_str
else:
static_dict['xgmi_plpd'] = policy_info
if 'profile' in current_platform_args:
if args.profile:
try:
profile_status = amdsmi_interface.amdsmi_get_gpu_power_profile_presets(args.gpu, 0)
# Parse available profiles from bitfield
available_profiles = self.helpers.parse_available_profiles(
profile_status['available_profiles']
)
# Get current profile name
current_profile = self.helpers.get_profile_name_from_mask(
profile_status['current']
)
# Store output
static_dict['profile'] = {
'available_profiles': available_profiles,
'current': current_profile,
'num_profiles': profile_status['num_profiles']
}
except amdsmi_exception.AmdSmiLibraryException as e:
static_dict['profile'] = e.get_error_info()
logging.debug("Failed to get power profile info for gpu %s | %s", gpu_id, e.get_error_info())
if 'process_isolation' in current_platform_args:
if args.process_isolation:
try:
status = amdsmi_interface.amdsmi_get_gpu_process_isolation(args.gpu)
status = "Enabled" if status else "Disabled"
except amdsmi_exception.AmdSmiLibraryException as e:
status = "N/A"
logging.debug("Failed to process isolation for gpu %s | %s", gpu_id, e.get_error_info())
static_dict['process_isolation'] = status
if 'numa' in current_platform_args:
if args.numa:
try:
numa_node_number = amdsmi_interface.amdsmi_topo_get_numa_node_number(args.gpu)
except amdsmi_exception.AmdSmiLibraryException as e:
numa_node_number = "N/A"
logging.debug("Failed to get numa node number for gpu %s | %s", gpu_id, e.get_error_info())
try:
numa_affinity = amdsmi_interface.amdsmi_get_gpu_topo_numa_affinity(args.gpu)
# -1 means No numa node is assigned to the GPU, so there is no numa affinity
if self.logger.is_human_readable_format() and numa_affinity == -1:
numa_affinity = "NONE"
except amdsmi_exception.AmdSmiLibraryException as e:
numa_affinity = "N/A"
logging.debug("Failed to get numa affinity for gpu %s | %s", gpu_id, e.get_error_info())
try:
cpu_set = amdsmi_interface.amdsmi_get_cpu_affinity_with_scope(args.gpu, amdsmi_interface.AmdSmiAffinityScope.NUMA_SCOPE)
cpu_set = [f"{cpus:016X}" for cpus in cpu_set]
cpu_set = {f'cpu_list_{i}': f"{cpus}" for i, cpus in enumerate(cpu_set)}
bitmask_ranges = self.helpers.get_bitmask_ranges(cpu_set)
cpu_affinity = {}
for key in cpu_set:
cpu_affinity[key] = {
"bitmask": cpu_set[key],
"cpu_cores_affinity" : bitmask_ranges[key]
}
except amdsmi_exception.AmdSmiLibraryException as e:
cpu_affinity = "N/A"
logging.debug("Failed to get cpu affinity for gpu %s | %s", gpu_id, e.get_error_info())
try:
socket_set = amdsmi_interface.amdsmi_get_cpu_affinity_with_scope(args.gpu, amdsmi_interface.AmdSmiAffinityScope.SOCKET_SCOPE)
socket_set = [f"{cpus:016X}" for cpus in socket_set]
socket_set = {f'cpu_list_{i}': f"{cpus}" for i, cpus in enumerate(socket_set)}
socket_bitmask_ranges = self.helpers.get_bitmask_ranges(socket_set)
socket_affinity = {}
for key in socket_set:
socket_affinity[key] = {
"bitmask": socket_set[key],
"cpu_cores_affinity": socket_bitmask_ranges.get(key, "N/A")
}
except amdsmi_exception.AmdSmiLibraryException as e:
socket_affinity = "N/A"
logging.debug("Failed to get socket affinity for gpu %s | %s", gpu_id, e.get_error_info())
static_dict['numa'] = { 'node' : numa_node_number,
'affinity' : numa_affinity,
'cpu_affinity' : cpu_affinity,
'socket_affinity' : socket_affinity}
if args.vram:
vram_info_dict = {"type" : "N/A",
"vendor" : "N/A",
"size" : "N/A",
"bit_width" : "N/A",
"max_bandwidth" : "N/A"}
try:
vram_info = amdsmi_interface.amdsmi_get_gpu_vram_info(args.gpu)
# Get vram type string
vram_type_enum = vram_info['vram_type']
if vram_type_enum == amdsmi_interface.amdsmi_wrapper.AMDSMI_VRAM_TYPE__MAX:
vram_type = "GDDR7"
else:
vram_type = amdsmi_interface.amdsmi_wrapper.amdsmi_vram_type_t__enumvalues[vram_type_enum]
# Remove amdsmi enum prefix
vram_type = vram_type.replace('AMDSMI_VRAM_TYPE_', '').replace('_', '')
# Get vram vendor string
vram_vendor = vram_info['vram_vendor']
if "PLACEHOLDER" in vram_vendor:
vram_vendor = "N/A"
# Assign cleaned values to vram_info_dict
vram_info_dict['type'] = vram_type
vram_info_dict['vendor'] = vram_vendor
# Populate vram size with unit
vram_info_dict['size'] = vram_info['vram_size']
vram_size_unit = "MB"
if self.logger.is_human_readable_format():
vram_info_dict['size'] = f"{vram_info['vram_size']} {vram_size_unit}"
if self.logger.is_json_format():
vram_info_dict['size'] = {"value" : vram_info['vram_size'],
"unit" : vram_size_unit}
# Populate bit width
vram_info_dict['bit_width'] = vram_info['vram_bit_width']
# Populate vram_max_bandwidth
vram_max_bw = vram_info['vram_max_bandwidth']
vram_max_bw_unit = 'GB/s'
if self.logger.is_human_readable_format():
vram_info_dict["max_bandwidth"] = f"{vram_max_bw} {vram_max_bw_unit if vram_max_bw != 'N/A' else ''}"
if self.logger.is_json_format():
vram_info_dict["max_bandwidth"] = {"value" : vram_max_bw,
"unit" : vram_max_bw_unit}
except amdsmi_exception.AmdSmiLibraryException as e:
logging.debug("Failed to get vram info for gpu %s | %s", gpu_id, e.get_error_info())
static_dict['vram'] = vram_info_dict
if args.cache:
try:
cache_info_list = amdsmi_interface.amdsmi_get_gpu_cache_info(args.gpu)['cache']
logging.debug(f"cache_info dictionary = {cache_info_list}")
for index, cache_info in enumerate(cache_info_list):
new_cache_info = {"cache" : index}
new_cache_info.update(cache_info)
cache_info_list[index] = new_cache_info
logging.debug(f"[after update] cache_info_list = {cache_info_list}")
cache_size_unit = "KB"
if self.logger.is_human_readable_format():
cache_info_dict_format = {}
for cache_dict in cache_info_list:
cache_index = "cache_" + str(cache_dict["cache"])
cache_info_dict_format[cache_index] = cache_dict
# Remove cache index from new dictionary
cache_info_dict_format[cache_index].pop("cache")
# Add cache_size unit
cache_size = f"{cache_info_dict_format[cache_index]['cache_size']} {cache_size_unit}"
cache_info_dict_format[cache_index]["cache_size"] = cache_size
# take cache_properties out of list -> display as string, removing brackets
cache_info_dict_format[cache_index]["cache_properties"] = ", ".join(cache_info_dict_format[cache_index]["cache_properties"])
cache_info_list = cache_info_dict_format
logging.debug(f"[human readable] cache_info_list = {cache_info_list}")
# Add cache_size_unit to json output
if self.logger.is_json_format():
for cache_dict in cache_info_list:
cache_dict["cache_size"] = {"value" : cache_dict["cache_size"],
"unit" : cache_size_unit}
except amdsmi_exception.AmdSmiLibraryException as e:
cache_info_list = "N/A"
logging.debug("Failed to get cache info for gpu %s | %s", gpu_id, e.get_error_info())
static_dict['cache_info'] = cache_info_list
# default to printing all clocks, if in current_platform_args; otherwise print specific clocks
if 'clock' in current_platform_args and (args.clock == True or isinstance(args.clock, list)):
original_clock_args = args.clock #save original args.clock value, so we can reset for multiple devices
if isinstance(args.clock, bool):
args.clock = ['sys', 'mem', 'df', 'soc', 'dcef', 'vclk0', 'vclk1', 'dclk0', 'dclk1']
if isinstance(args.clock, list):
# remove potential duplicates from list
args.clock = list(set(args.clock))
# check that clock is valid option
if "all" in args.clock or len(args.clock) == 0:
args.clock = ['sys', 'mem', 'df', 'soc', 'dcef', 'vclk0', 'vclk1', 'dclk0', 'dclk1']
clk_dict = {
'sys': "N/A",
'mem': "N/A",
'df': "N/A",
'soc': "N/A",
'dcef': "N/A",
'vclk0': "N/A",
'vclk1': "N/A",
'dclk0': "N/A",
'dclk1': "N/A",
}
for clk in list(clk_dict.keys()):
if clk not in args.clock:
del clk_dict[clk]
for clk in args.clock:
if clk in self.convert_clock_type:
clk_type_conversion = self.convert_clock_type[clk]
else:
clk_type_conversion = "N/A"
output_format = self.helpers.get_output_format()
raise AmdSmiInvalidParameterException('static', clk_type, output_format) # clk type given is bad
try:
frequencies = amdsmi_interface.amdsmi_get_clk_freq(args.gpu, clk_type_conversion)
# some clocks may have a sysfs file but no frequencies for whatever reason.
if len(frequencies['frequency']) == 0:
freq_dict = "N/A"
continue
freq_dict = {}
current_level = frequencies['current']
freq_dict.update({'current_level':current_level})
current_frequency = str(self.helpers.convert_SI_unit(frequencies['frequency'][current_level], AMDSMIHelpers.SI_Unit.MICRO)) + "MHz"
freq_dict.update({'current_frequency':current_frequency})
freq_dict.update({'frequency_levels':{}})
if frequencies["num_supported"] != 0:
for level in range(len(frequencies['frequency'])):
if frequencies['frequency'][level] != "N/A":
freq = str(self.helpers.convert_SI_unit(frequencies['frequency'][level], AMDSMIHelpers.SI_Unit.MICRO)) + " MHz"
freq_dict['frequency_levels'].update({f"Level {level}":freq})
else:
freq_dict['frequency_levels'].update({f"Level {level}":"N/A"})
else:
freq_dict = "N/A"
except amdsmi_exception.AmdSmiLibraryException as e:
freq_dict = "N/A"
logging.debug("Failed to get clock info for gpu %s | %s", gpu_id, e.get_error_info())
clk_dict[clk] = freq_dict
static_dict['clock'] = clk_dict
else:
raise amdsmi_exception.AmdSmiParameterException(args.clock, 'list[str]')
# if original_clock_args is a boolean, set it back to the original value
if isinstance(original_clock_args, bool):
args.clock = original_clock_args
# Convert and store output by pid for csv format
multiple_devices_csv_override = False
if self.logger.is_csv_format():
# For NUMA data - flatten CPU affinity lists
if 'numa' in static_dict and isinstance(static_dict['numa'], dict):
numa_data = static_dict.pop('numa')
multiple_devices_csv_override = True
# Get data
node = numa_data.get('node', 'N/A')
affinity = numa_data.get('affinity', 'N/A')
cpu_affinity = numa_data.get('cpu_affinity', {})
socket_affinity = numa_data.get('socket_affinity', {})
# Create a flattened row for list entry
row_dict = static_dict.copy()
if cpu_affinity and isinstance(cpu_affinity, dict):
for cpu_list_key in cpu_affinity.keys():
cpu_entry = cpu_affinity[cpu_list_key]
socket_entry = socket_affinity.get(cpu_list_key, {"bitmask": "N/A", "cpu_cores_affinity": "N/A"})
row_dict.update({
'node': node,
'affinity': affinity,
'cpu_list': cpu_list_key,
'bitmask': cpu_entry.get('bitmask'),
'cpu_cores_affinity': cpu_entry.get('cpu_cores_affinity'),
'socket_bitmask': socket_entry.get('bitmask'),
'socket_cpu_cores_affinity': socket_entry.get('cpu_cores_affinity')
})
self.logger.store_output(args.gpu, 'values', row_dict)
self.logger.store_gpu_json_output.append(row_dict)
self.logger.store_multiple_device_output()
else:
row_dict.update({
'node': node,
'affinity': affinity,
'cpu_list': 'N/A',
'bitmask': 'N/A',
'cpu_cores_affinity': 'N/A',
'socket_bitmask': 'N/A',
'socket_cpu_cores_affinity': 'N/A'
})
self.logger.store_output(args.gpu, 'values', row_dict)
self.logger.store_gpu_json_output.append(row_dict)
# expand if ras blocks are populated
elif self.helpers.is_linux() and self.helpers.is_baremetal() and args.ras:
if isinstance(static_dict['ras']['ecc_block_state'], list):
ecc_block_dicts = static_dict['ras'].pop('ecc_block_state')
multiple_devices_csv_override = True
for ecc_block_dict in ecc_block_dicts:
for key, value in ecc_block_dict.items():
self.logger.store_output(args.gpu, key, value)
self.logger.store_output(args.gpu, 'values', static_dict)
self.logger.store_gpu_json_output.append(static_dict)
self.logger.store_multiple_device_output()
else:
# Store values if ras has an error
self.logger.store_output(args.gpu, 'values', static_dict)
self.logger.store_gpu_json_output.append(static_dict)
if self.helpers.is_linux() and self.helpers.is_virtual_os():
self.logger.store_output(args.gpu, 'values', static_dict)
self.logger.store_gpu_json_output.append(static_dict)
else:
self.logger.store_output(args.gpu, 'values', static_dict)
self.logger.store_gpu_json_output.append(static_dict)
elif self.logger.is_json_format():
self.logger.store_gpu_json_output.append(static_dict)
else:
# Store values in logger.output
self.logger.store_output(args.gpu, 'values', static_dict)
if multiple_devices:
self.logger.store_multiple_device_output()
return # Skip printing when there are multiple devices
if not self.logger.is_json_format():
self.logger.print_output(multiple_device_enabled=multiple_devices_csv_override)
def static(self, args, multiple_devices=False, gpu=None, asic=None,
bus=None, vbios=None, limit=None, driver=None, ras=None,
board=None, numa=None, vram=None, cache=None, partition=None,
dfc_ucode=None, fb_info=None, num_vf=None, cpu=None,
interface_ver=None, soc_pstate=None, xgmi_plpd = None, process_isolation=None,
clock=None, profile=None):
"""Get Static information for target gpu and cpu
Args:
args (Namespace): Namespace containing the parsed CLI args
multiple_devices (bool, optional): True if checking for multiple devices. Defaults to False.
gpu (device_handle, optional): device_handle for target device. Defaults to None.
asic (bool, optional): Value override for args.asic. Defaults to None.
bus (bool, optional): Value override for args.bus. Defaults to None.
vbios (bool, optional): Value override for args.vbios. Defaults to None.
limit (bool, optional): Value override for args.limit. Defaults to None.
driver (bool, optional): Value override for args.driver. Defaults to None.
ras (bool, optional): Value override for args.ras. Defaults to None.
board (bool, optional): Value override for args.board. Defaults to None.
numa (bool, optional): Value override for args.numa. Defaults to None.
vram (bool, optional): Value override for args.vram. Defaults to None.
cache (bool, optional): Value override for args.cache. Defaults to None.
partition (bool, optional): Value override for args.partition. Defaults to None.
dfc_ucode (bool, optional): Value override for args.dfc_ucode. Defaults to None.
fb_info (bool, optional): Value override for args.fb_info. Defaults to None.
num_vf (bool, optional): Value override for args.num_vf. Defaults to None.
cpu (cpu_handle, optional): cpu_handle for target device. Defaults to None.
interface_ver (bool, optional): Value override for args.interface_ver. Defaults to None
soc_pstate (bool, optional): Value override for args.soc_pstate. Defaults to None.
xgmi_plpd (bool, optional): Value override for args.xgmi_plpd. Defaults to None.
process_isolation (bool, optional): Value override for args.process_isolation. Defaults to None.
Raises:
IndexError: Index error if gpu list is empty
Returns:
None: Print output via AMDSMILogger to destination
"""
# Mutually exclusive arguments
if cpu:
args.cpu = cpu
if gpu:
args.gpu = gpu
# Check if a CPU argument has been set
cpu_args_enabled = False
cpu_attributes = ["smu", "interface_ver"]
for attr in cpu_attributes:
if hasattr(args, attr):
if getattr(args, attr):
cpu_args_enabled = True
break
# Check if a GPU argument has been set
gpu_args_enabled = False
gpu_attributes = ["asic", "bus", "vbios", "limit", "driver", "ras",
"board", "numa", "vram", "cache", "partition",
"dfc_ucode", "fb_info", "num_vf", "soc_pstate", "xgmi_plpd",
"process_isolation", "clock", "profile"]
for attr in gpu_attributes:
if hasattr(args, attr):
if getattr(args, attr):
gpu_args_enabled = True
break
# Handle CPU and GPU intialization cases
if self.helpers.is_amd_hsmp_initialized() and self.helpers.is_amdgpu_initialized():
# Print out all CPU and all GPU static info only if no device was specified.
# If a GPU or CPU argument is provided only print out the specified device.
if args.cpu == None and args.gpu == None:
if not cpu_args_enabled and not gpu_args_enabled:
args.cpu = self.cpu_handles
args.gpu = self.device_handles
# Handle cases where the user has only specified an argument and no specific device
if args.gpu == None and gpu_args_enabled:
args.gpu = self.device_handles
if args.cpu == None and cpu_args_enabled:
args.cpu = self.cpu_handles
if args.cpu:
self.static_cpu(args, multiple_devices, cpu, interface_ver)
if args.gpu:
self.logger.output = {}
self.logger.clear_multiple_devices_output()
self.static_gpu(args, multiple_devices, gpu, asic,
bus, vbios, limit, driver, ras,
board, numa, vram, cache, partition,
dfc_ucode, fb_info, num_vf, soc_pstate, xgmi_plpd,
process_isolation, clock, profile)
elif self.helpers.is_amd_hsmp_initialized(): # Only CPU is initialized
if args.cpu == None:
args.cpu = self.cpu_handles
self.static_cpu(args, multiple_devices, cpu, interface_ver)
elif self.helpers.is_amdgpu_initialized(): # Only GPU is initialized
if args.gpu == None:
args.gpu = self.device_handles
self.logger.clear_multiple_devices_output()
self.static_gpu(args, multiple_devices, gpu, asic,
bus, vbios, limit, driver, ras,
board, numa, vram, cache, partition,
dfc_ucode, fb_info, num_vf, soc_pstate, xgmi_plpd,
process_isolation, clock, profile)
if self.logger.is_json_format():
self.logger.combine_arrays_to_json()
def firmware(self, args, multiple_devices=False, gpu=None, fw_list=True):
""" Get Firmware information for target gpu
Args:
args (Namespace): Namespace containing the parsed CLI args
multiple_devices (bool, optional): True if checking for multiple devices. Defaults to False.
gpu (device_handle, optional): device_handle for target device. Defaults to None.
fw_list (bool, optional): True to get list of all firmware information
Raises:
IndexError: Index error if gpu list is empty
Returns:
None: Print output via AMDSMILogger to destination
"""
if gpu:
args.gpu = gpu
if fw_list:
args.fw_list = fw_list
# Handle No GPU passed
if args.gpu == None:
args.gpu = self.device_handles
# Handle multiple GPUs
handled_multiple_gpus, device_handle = self.helpers.handle_gpus(args, self.logger, self.firmware)
if handled_multiple_gpus:
return # This function is recursive
args.gpu = device_handle
fw_list = {}
# Get gpu_id for logging
gpu_id = self.helpers.get_gpu_id_from_device_handle(args.gpu)
if args.fw_list:
try:
fw_info = amdsmi_interface.amdsmi_get_fw_info(args.gpu)
for fw_index, fw_entry in enumerate(fw_info['fw_list']):
# Change fw_name to fw_id
fw_entry['fw_id'] = fw_entry.pop('fw_name').name.replace("AMDSMI_FW_ID_", "")
fw_entry['fw_version'] = fw_entry.pop('fw_version') # popping to ensure order
# Add custom human readable formatting
if self.logger.is_human_readable_format():
fw_info['fw_list'][fw_index] = {f'FW {fw_index}': fw_entry}
else:
fw_info['fw_list'][fw_index] = fw_entry
fw_list.update(fw_info)
except amdsmi_exception.AmdSmiLibraryException as e:
fw_list['fw_list'] = "N/A"
logging.debug("Failed to get firmware info for gpu %s | %s", gpu_id, e.get_error_info())
multiple_devices_csv_override = False
# Convert and store output by pid for csv format
if self.logger.is_csv_format():
fw_key = 'fw_list'
for fw_info_dict in fw_list[fw_key]:
for key, value in fw_info_dict.items():
multiple_devices_csv_override = True
self.logger.store_output(args.gpu, key, value)
self.logger.store_multiple_device_output()
else:
# Store values in logger.output
self.logger.store_output(args.gpu, 'values', fw_list)
if multiple_devices:
self.logger.store_multiple_device_output()
return # Skip printing when there are multiple devices
self.logger.print_output(multiple_device_enabled=multiple_devices_csv_override)
def bad_pages(self, args, multiple_devices=False, gpu=None, retired=None, pending=None, un_res=None, hex_format=None):
""" Get bad pages information for target gpu
Args:
args (Namespace): Namespace containing the parsed CLI args
multiple_devices (bool, optional): True if checking for multiple devices. Defaults to False.
gpu (device_handle, optional): device_handle for target device. Defaults to None.
retired (bool, optional) - Value override for args.retired
pending (bool, optional) - Value override for args.pending/
un_res (bool, optional) - Value override for args.un_res
hex_format (bool, optional) - Value override for args.hex
Raises:
IndexError: Index error if gpu list is empty
Returns:
None: Print output via AMDSMILogger to destination
"""
# Set args.* to passed in arguments
if gpu:
args.gpu = gpu
if retired:
args.retired = retired
if pending:
args.pending = pending
if un_res:
args.un_res = un_res
if hex_format is not None:
args.hex = hex_format
# Handle No GPU passed
if args.gpu == None:
args.gpu = self.device_handles
# Handle multiple GPUs
handled_multiple_gpus, device_handle = self.helpers.handle_gpus(args, self.logger, self.bad_pages)
if handled_multiple_gpus:
return # This function is recursive
args.gpu = device_handle
# If all arguments are False, the print all bad_page information
if not any([args.retired, args.pending, args.un_res]):
args.retired = args.pending = args.un_res = True
values_dict = {}
# Get gpu_id for logging
gpu_id = self.helpers.get_gpu_id_from_device_handle(args.gpu)
bad_pages_not_found = "No bad pages found."
try:
bad_page_info = amdsmi_interface.amdsmi_get_gpu_bad_page_info(args.gpu)
# If bad_page_info is an empty list overwrite with not found error statement
if bad_page_info == []:
bad_page_info = bad_pages_not_found
bad_page_error = True
else:
bad_page_error = False
except amdsmi_exception.AmdSmiLibraryException as e:
bad_page_info = "N/A"
bad_page_error = True
logging.debug("Failed to get bad page info for gpu %s | %s", gpu_id, e.get_error_info())
if args.retired:
if bad_page_error:
values_dict['retired'] = bad_page_info
else:
bad_page_info_output = []
for bad_page in bad_page_info:
if bad_page["status"] == amdsmi_interface.AmdSmiMemoryPageStatus.RESERVED:
bad_page_info_entry = {}
# Format page address and size based on --hex flag
if args.hex:
bad_page_info_entry["page_address"] = f"0x{bad_page['page_address']:x}"
bad_page_info_entry["page_size"] = f"0x{bad_page['page_size']:x}"
else:
bad_page_info_entry["page_address"] = bad_page["page_address"]
bad_page_info_entry["page_size"] = bad_page["page_size"]
status_string = amdsmi_interface.amdsmi_wrapper.amdsmi_memory_page_status_t__enumvalues[bad_page["status"]]
bad_page_info_entry["status"] = status_string.replace("AMDSMI_MEM_PAGE_STATUS_", "")
bad_page_info_output.append(bad_page_info_entry)
# Remove brackets if there is only one value
if len(bad_page_info_output) == 1:
bad_page_info_output = bad_page_info_output[0]
if bad_page_info_output == []:
values_dict['retired'] = bad_pages_not_found
else:
values_dict['retired'] = bad_page_info_output
if args.pending:
if bad_page_error:
values_dict['pending'] = bad_page_info
else:
bad_page_info_output = []
for bad_page in bad_page_info:
if bad_page["status"] == amdsmi_interface.AmdSmiMemoryPageStatus.PENDING:
bad_page_info_entry = {}
# Format page address and size based on --hex flag
if args.hex:
bad_page_info_entry["page_address"] = f"0x{bad_page['page_address']:x}"
bad_page_info_entry["page_size"] = f"0x{bad_page['page_size']:x}"
else:
bad_page_info_entry["page_address"] = bad_page["page_address"]
bad_page_info_entry["page_size"] = bad_page["page_size"]
status_string = amdsmi_interface.amdsmi_wrapper.amdsmi_memory_page_status_t__enumvalues[bad_page["status"]]
bad_page_info_entry["status"] = status_string.replace("AMDSMI_MEM_PAGE_STATUS_", "")
bad_page_info_output.append(bad_page_info_entry)
# Remove brackets if there is only one value
if len(bad_page_info_output) == 1:
bad_page_info_output = bad_page_info_output[0]
if bad_page_info_output == []:
values_dict['pending'] = bad_pages_not_found
else:
values_dict['pending'] = bad_page_info_output
if args.un_res:
if bad_page_error:
values_dict['un_res'] = bad_page_info
else:
bad_page_info_output = []
for bad_page in bad_page_info:
if bad_page["status"] == amdsmi_interface.AmdSmiMemoryPageStatus.UNRESERVABLE:
bad_page_info_entry = {}
# Format page address and size based on --hex flag
if hasattr(args, 'hex') and args.hex:
bad_page_info_entry["page_address"] = f"0x{bad_page['page_address']:x}"
bad_page_info_entry["page_size"] = f"0x{bad_page['page_size']:x}"
else:
bad_page_info_entry["page_address"] = bad_page["page_address"]
bad_page_info_entry["page_size"] = bad_page["page_size"]
status_string = amdsmi_interface.amdsmi_wrapper.amdsmi_memory_page_status_t__enumvalues[bad_page["status"]]
bad_page_info_entry["status"] = status_string.replace("AMDSMI_MEM_PAGE_STATUS_", "")
bad_page_info_output.append(bad_page_info_entry)
# Remove brackets if there is only one value
if len(bad_page_info_output) == 1:
bad_page_info_output = bad_page_info_output[0]
if bad_page_info_output == []:
values_dict['un_res'] = bad_pages_not_found
else:
values_dict['un_res'] = bad_page_info_output
# Store values in logger.output
self.logger.store_output(args.gpu, 'values', values_dict)
if multiple_devices:
self.logger.store_multiple_device_output()
return # Skip printing when there are multiple devices
self.logger.print_output()
def metric_gpu(self, args, multiple_devices=False, watching_output=False, gpu=None,
usage=None, watch=None, watch_time=None, iterations=None, power=None,
clock=None, temperature=None, ecc=None, ecc_blocks=None, pcie=None,
fan=None, voltage_curve=None, overdrive=None, perf_level=None,
xgmi_err=None, energy=None, mem_usage=None, voltage=None, schedule=None,
guard=None, guest_data=None, fb_usage=None, xgmi=None, throttle=None,
base_board=None, gpu_board=None):
"""Get Metric information for target gpu
Args:
args (Namespace): Namespace containing the parsed CLI args
multiple_devices (bool, optional): True if checking for multiple devices. Defaults to False.
watching_output (bool, optional): True if watch argument has been set. Defaults to False.
gpu (device_handle, optional): device_handle for target device. Defaults to None.
usage (bool, optional): Value override for args.usage. Defaults to None.
watch (Positive int, optional): Value override for args.watch. Defaults to None.
watch_time (Positive int, optional): Value override for args.watch_time. Defaults to None.
iterations (Positive int, optional): Value override for args.iterations. Defaults to None.
power (bool, optional): Value override for args.power. Defaults to None.
clock (bool, optional): Value override for args.clock. Defaults to None.
temperature (bool, optional): Value override for args.temperature. Defaults to None.
ecc (bool, optional): Value override for args.ecc. Defaults to None.
ecc_blocks (bool, optional): Value override for args.ecc. Defaults to None.
pcie (bool, optional): Value override for args.pcie. Defaults to None.
fan (bool, optional): Value override for args.fan. Defaults to None.
voltage_curve (bool, optional): Value override for args.voltage_curve. Defaults to None.
overdrive (bool, optional): Value override for args.overdrive. Defaults to None.
perf_level (bool, optional): Value override for args.perf_level. Defaults to None.
xgmi_err (bool, optional): Value override for args.xgmi_err. Defaults to None.
energy (bool, optional): Value override for args.energy. Defaults to None.
mem_usage (bool, optional): Value override for args.mem_usage. Defaults to None.
voltage (bool, optional): Value override for args.voltage. Defaults to None.
schedule (bool, optional): Value override for args.schedule. Defaults to None.
guard (bool, optional): Value override for args.guard. Defaults to None.
guest_data (bool, optional): Value override for args.guest_data. Defaults to None.
fb_usage (bool, optional): Value override for args.fb_usage. Defaults to None.
xgmi (bool, optional): Value override for args.xgmi. Defaults to None.
throttle (bool, optional): Value override for args.throttle. Defaults to None.
Raises:
IndexError: Index error if gpu list is empty
Returns:
None: Print output via AMDSMILogger to destination
"""
# Set args.* to passed in arguments
if gpu:
args.gpu = gpu
if watch:
args.watch = watch
if watch_time:
args.watch_time = watch_time
if iterations:
args.iterations = iterations
# Store args that are applicable to the current platform
current_platform_args = []
current_platform_values = []
if not self.helpers.is_hypervisor() and not self.helpers.is_windows():
if mem_usage:
args.mem_usage = mem_usage
current_platform_args += ["mem_usage"]
current_platform_values += [args.mem_usage]
if self.helpers.is_hypervisor() or self.helpers.is_baremetal() or self.helpers.is_linux():
if usage:
args.usage = usage
if base_board:
args.base_board = base_board
if gpu_board:
args.gpu_board = gpu_board
if power:
args.power = power
if clock:
args.clock = clock
if temperature:
args.temperature = temperature
if voltage:
args.voltage = voltage
if pcie:
args.pcie = pcie
if ecc:
args.ecc = ecc
if ecc_blocks:
args.ecc_blocks = ecc_blocks
current_platform_args += ["usage", "power", "clock", "temperature", "voltage", "pcie", "ecc", "ecc_blocks", "base_board","gpu_board"]
current_platform_values += [args.usage, args.power, args.clock,
args.temperature, args.voltage, args.pcie]
current_platform_values += [args.ecc, args.ecc_blocks, args.base_board, args.gpu_board]
if self.helpers.is_baremetal() and self.helpers.is_linux():
if fan:
args.fan = fan
if voltage_curve:
args.voltage_curve = voltage_curve
if overdrive:
args.overdrive = overdrive
if perf_level:
args.perf_level = perf_level
if xgmi_err:
args.xgmi_err = xgmi_err
if energy:
args.energy = energy
if throttle:
args.violation = throttle
args.throttle = throttle
current_platform_args += ["fan", "voltage_curve", "overdrive", "perf_level",
"xgmi_err", "energy", "throttle"]
current_platform_values += [args.fan, args.voltage_curve, args.overdrive,
args.perf_level, args.xgmi_err, args.energy, args.throttle,
]
if self.helpers.is_hypervisor():
if schedule:
args.schedule = schedule
if guard:
args.guard = guard
if guest_data:
args.guest_data = guest_data
if fb_usage:
args.fb_usage = fb_usage
if xgmi:
args.xgmi = xgmi
current_platform_args += ["schedule", "guard", "guest_data", "fb_usage", "xgmi"]
current_platform_values += [args.schedule, args.guard, args.guest_data,
args.fb_usage, args.xgmi]
# Handle No GPU passed
if args.gpu == None:
args.gpu = self.device_handles
if not self.group_check_printed:
self.helpers.check_required_groups(check_render=True, check_video=False)
self.group_check_printed = True
# Handle watch logic, will only enter this block once
if args.watch:
self.helpers.handle_watch(args=args, subcommand=self.metric_gpu, logger=self.logger)
return
# Handle multiple GPUs
if isinstance(args.gpu, list):
if len(args.gpu) > 1:
# Deepcopy gpus as recursion will destroy the gpu list
stored_gpus = []
for gpu in args.gpu:
stored_gpus.append(gpu)
# Store output from multiple devices
for device_handle in args.gpu:
self.metric_gpu(args, multiple_devices=True, watching_output=watching_output, gpu=device_handle)
# Reload original gpus
args.gpu = stored_gpus
# Print multiple device output
if not self.logger.is_json_format():
self.logger.print_output(multiple_device_enabled=True, watching_output=watching_output)
# Add output to total watch output and clear multiple device output
if watching_output:
self.logger.store_watch_output(multiple_device_enabled=True)
# Flush the watching output
self.logger.print_output(multiple_device_enabled=True, watching_output=watching_output)
return
elif len(args.gpu) == 1:
args.gpu = args.gpu[0]
else:
raise IndexError("args.gpu should not be an empty list")
# Get gpu_id for logging
gpu_id = self.helpers.get_gpu_id_from_device_handle(args.gpu)
if args.loglevel == "DEBUG":
try:
# Get GPU Metrics table version
gpu_metric_version_info = amdsmi_interface.amdsmi_get_gpu_metrics_header_info(args.gpu)
gpu_metric_version_str = json.dumps(gpu_metric_version_info, indent=4)
logging.debug("GPU Metrics table Version for GPU %s | %s", gpu_id, gpu_metric_version_str)
except amdsmi_exception.AmdSmiLibraryException as e:
logging.debug("#1 - Unable to load GPU Metrics table version for %s | %s", gpu_id, e.get_error_info())
try:
# Get GPU Metrics table
gpu_metric_debug_info = amdsmi_interface.amdsmi_get_gpu_metrics_info(args.gpu)
gpu_metric_str = json.dumps(gpu_metric_debug_info, indent=4)
logging.debug("GPU Metrics table for GPU %s | %s", gpu_id, str(gpu_metric_str))
except amdsmi_exception.AmdSmiLibraryException as e:
logging.debug("#2 - Unable to load GPU Metrics table for %s | %s", gpu_id, e.get_error_info())
logging.debug(f"Metric Arg information for GPU {gpu_id} on {self.helpers.os_info()}")
logging.debug(f"Args: {current_platform_args}")
logging.debug(f"Values: {current_platform_values}")
# Set the platform applicable args to True if no args are set
if not any(current_platform_values):
for arg in current_platform_args:
setattr(args, arg, True)
# Add timestamp and store values for specified arguments
values_dict = {}
is_partition_metrics = False # True if we get the metrics from xcp_metrics file (amdsmi_get_gpu_partition_metrics_info)
#get metric info only once per gpu, this will speed up data output
try:
# Get GPU Metrics table
gpu_metric = amdsmi_interface.amdsmi_get_gpu_metrics_info(args.gpu)
except amdsmi_exception.AmdSmiLibraryException as e:
logging.debug("#3 - Unable to load GPU Metrics table for %s | %s", gpu_id, e.get_error_info())
gpu_metric = amdsmi_interface._NA_amdsmi_get_gpu_metrics_info()
# Workaround for XCP (partition) metrics not providing num_partition in v1.9+/v1.1+
# Provides original formatting for earlier metric versions
partition_metric_info = self.helpers._get_metric_version_and_partition_info(gpu_metric, is_partition_metrics, gpu_id, args.gpu)
num_partition = partition_metric_info['num_partition']
if self.logger.is_json_format():
values_dict['gpu'] = int(gpu_id)
# Populate the pcie_dict first due to multiple gpu metrics calls incorrectly increasing bandwidth
if "pcie" in current_platform_args:
if args.pcie:
pcie_dict = {"width": "N/A",
"speed": "N/A",
"bandwidth": "N/A",
"replay_count" : "N/A",
"l0_to_recovery_count" : "N/A",
"replay_roll_over_count" : "N/A",
"nak_sent_count" : "N/A",
"nak_received_count" : "N/A",
"current_bandwidth_sent": "N/A",
"current_bandwidth_received": "N/A",
"max_packet_size": "N/A",
"lc_perf_other_end_recovery": "N/A"}
try:
pcie_metric = amdsmi_interface.amdsmi_get_pcie_info(args.gpu)['pcie_metric']
logging.debug("PCIE Metric for %s | %s", gpu_id, pcie_metric)
pcie_dict['width'] = pcie_metric['pcie_width']
if pcie_metric['pcie_speed'] != "N/A":
if pcie_metric['pcie_speed'] % 1000 != 0:
pcie_speed_GTs_value = round(pcie_metric['pcie_speed'] / 1000, 1)
else:
pcie_speed_GTs_value = round(pcie_metric['pcie_speed'] / 1000)
pcie_dict['speed'] = pcie_speed_GTs_value
pcie_dict['bandwidth'] = pcie_metric['pcie_bandwidth']
pcie_dict['replay_count'] = pcie_metric['pcie_replay_count']
if pcie_dict['replay_count'] == "N/A":
try:
pcie_replay = amdsmi_interface.amdsmi_get_gpu_pci_replay_counter(args.gpu)
pcie_dict['replay_count'] = pcie_replay
except amdsmi_exception.AmdSmiLibraryException as e:
logging.debug("Failed to get sysfs pcie replay counter on gpu %s | %s", gpu_id, e.get_error_info())
pcie_dict['l0_to_recovery_count'] = pcie_metric['pcie_l0_to_recovery_count']
pcie_dict['replay_roll_over_count'] = pcie_metric['pcie_replay_roll_over_count']
pcie_dict['nak_received_count'] = pcie_metric['pcie_nak_received_count']
pcie_dict['nak_sent_count'] = pcie_metric['pcie_nak_sent_count']
pcie_dict['lc_perf_other_end_recovery'] = pcie_metric['pcie_lc_perf_other_end_recovery_count']
pcie_speed_unit = 'GT/s'
pcie_bw_unit = 'Mb/s'
if self.logger.is_human_readable_format():
if pcie_dict['speed'] != "N/A":
pcie_dict['speed'] = f"{pcie_dict['speed']} {pcie_speed_unit}"
if pcie_dict['bandwidth'] != "N/A":
pcie_dict['bandwidth'] = f"{pcie_dict['bandwidth']} {pcie_bw_unit}"
if self.logger.is_json_format():
if pcie_dict['speed'] != "N/A":
pcie_dict['speed'] = {"value" : pcie_dict['speed'],
"unit" : pcie_speed_unit}
if pcie_dict['bandwidth'] != "N/A":
pcie_dict['bandwidth'] = {"value" : pcie_dict['bandwidth'],
"unit" : pcie_bw_unit}
except amdsmi_exception.AmdSmiLibraryException as e:
logging.debug("Failed to get pcie link status for gpu %s | %s", gpu_id, e.get_error_info())
try:
pcie_bw = amdsmi_interface.amdsmi_get_gpu_pci_throughput(args.gpu)
sent = pcie_bw['sent'] * pcie_bw['max_pkt_sz']
received = pcie_bw['received'] * pcie_bw['max_pkt_sz']
bw_unit = "Mb/s"
packet_size_unit = "B"
if sent > 0:
sent = sent // 1024 // 1024
if received > 0:
received = received // 1024 // 1024
if self.logger.is_human_readable_format():
sent = f"{sent} {bw_unit}"
received = f"{received} {bw_unit}"
pcie_bw['max_pkt_sz'] = f"{pcie_bw['max_pkt_sz']} {packet_size_unit}"
if self.logger.is_json_format():
sent = {"value" : sent,
"unit" : bw_unit}
received = {"value" : received,
"unit" : bw_unit}
pcie_bw['max_pkt_sz'] = {"value" : pcie_bw['max_pkt_sz'],
"unit" : packet_size_unit}
pcie_dict['current_bandwidth_sent'] = sent
pcie_dict['current_bandwidth_received'] = received
pcie_dict['max_packet_size'] = pcie_bw['max_pkt_sz']
except amdsmi_exception.AmdSmiLibraryException as e:
logging.debug("Failed to get pcie bandwidth for gpu %s | %s", gpu_id, e.get_error_info())
if "usage" in current_platform_args:
if args.usage:
try:
engine_usage = amdsmi_interface.amdsmi_get_gpu_activity(args.gpu)
logging.debug(f"engine_usage dictionary = {engine_usage}")
# TODO: move vcn_activity and jpeg_activity into amdsmi_get_gpu_activity
engine_usage['vcn_activity'] = gpu_metric['vcn_activity']
engine_usage['jpeg_activity'] = gpu_metric['jpeg_activity']
engine_usage['gfx_busy_inst'] = "N/A"
engine_usage['jpeg_busy'] = "N/A"
engine_usage['vcn_busy'] = "N/A"
if num_partition != "N/A":
# these are one after another, in order to display each in sub-sections
new_xcp_dict = {}
for current_xcp in range(num_partition):
new_xcp_dict[f"xcp_{current_xcp}"] = gpu_metric['xcp_stats.gfx_busy_inst'][current_xcp]
engine_usage['gfx_busy_inst'] = new_xcp_dict
new_xcp_dict = {}
for current_xcp in range(num_partition):
new_xcp_dict[f"xcp_{current_xcp}"] = gpu_metric['xcp_stats.jpeg_busy'][current_xcp]
engine_usage['jpeg_busy'] = new_xcp_dict
new_xcp_dict = {}
for current_xcp in range(num_partition):
new_xcp_dict[f"xcp_{current_xcp}"] = gpu_metric['xcp_stats.vcn_busy'][current_xcp]
engine_usage['vcn_busy'] = new_xcp_dict
logging.debug(f"After updates to engine_usage dictionary = {engine_usage}")
for key, value in engine_usage.items():
activity_unit = '%'
if self.logger.is_human_readable_format():
if isinstance(value, list):
for index, activity in enumerate(value):
if activity != "N/A":
engine_usage[key][index] = f"{activity} {activity_unit}"
# Convert list to a string for human readable format
engine_usage[key] = '[' + ", ".join(engine_usage[key]) + ']'
elif isinstance(value, dict):
for k, v in value.items():
for index, activity in enumerate(v):
if activity != "N/A":
value[k][index] = f"{activity} {activity_unit}"
# Convert list to a string for human readable format
value[k] = '[' + ", ".join(value[k]) + ']'
elif value != "N/A":
engine_usage[key] = f"{value} {activity_unit}"
if self.logger.is_json_format():
if isinstance(value, list):
for index, activity in enumerate(value):
if activity != "N/A":
engine_usage[key][index] = {"value" : activity,
"unit" : activity_unit}
elif isinstance(value, dict):
for k, v in value.items():
for index, activity in enumerate(v):
if activity != "N/A":
value[k][index] = {"value" : activity,
"unit" : activity_unit}
elif value != "N/A":
engine_usage[key] = {"value" : value,
"unit" : activity_unit}
values_dict['usage'] = engine_usage
except Exception as e:
values_dict['usage'] = "N/A"
logging.debug("Failed to get gpu activity for gpu %s | %s", gpu_id, e)
if "power" in current_platform_args:
if args.power:
power_dict = {'socket_power': "N/A",
'gfx_voltage': "N/A",
'soc_voltage': "N/A",
'mem_voltage': "N/A",
'throttle_status': "N/A",
'power_management': "N/A"}
try:
voltage_unit = "mV"
power_unit = "W"
power_info = amdsmi_interface.amdsmi_get_power_info(args.gpu)
for key, value in power_info.items():
if "voltage" in key:
power_info[key] = self.helpers.unit_format(self.logger,
value,
voltage_unit)
elif 'power' in key:
power_info[key] = self.helpers.unit_format(self.logger,
value,
power_unit)
power_dict['socket_power'] = power_info['socket_power']
power_dict['gfx_voltage'] = power_info['gfx_voltage']
power_dict['soc_voltage'] = power_info['soc_voltage']
power_dict['mem_voltage'] = power_info['mem_voltage']
except amdsmi_exception.AmdSmiLibraryException as e:
logging.debug("Failed to get power info for gpu %s | %s", gpu_id, e.get_error_info())
try:
is_power_management_enabled = amdsmi_interface.amdsmi_is_gpu_power_management_enabled(args.gpu)
if is_power_management_enabled:
power_dict['power_management'] = "ENABLED"
else:
power_dict['power_management'] = "DISABLED"
except amdsmi_exception.AmdSmiLibraryException as e:
logging.debug("Failed to get power management status for gpu %s | %s", gpu_id, e.get_error_info())
try:
power_dict['throttle_status'] = "N/A"
throttle_status = gpu_metric['throttle_status']
if throttle_status != "N/A":
if throttle_status:
power_dict['throttle_status'] = "THROTTLED"
else:
power_dict['throttle_status'] = "UNTHROTTLED"
except Exception as e:
logging.debug("Failed to get throttle status for gpu %s | %s", gpu_id, e)
values_dict['power'] = power_dict
if "clock" in current_platform_args:
if args.clock:
# Populate Skeleton output with N/A
clocks = {}
for clock_index in range(amdsmi_interface.AMDSMI_MAX_NUM_GFX_CLKS):
gfx_index = f"gfx_{clock_index}"
clocks[gfx_index] = {"clk" : "N/A",
"min_clk" : "N/A",
"max_clk" : "N/A",
"clk_locked" : "N/A",
"deep_sleep" : "N/A"}
clocks["mem_0"] = {"clk" : "N/A",
"min_clk" : "N/A",
"max_clk" : "N/A",
"clk_locked" : "N/A",
"deep_sleep" : "N/A"}
for clock_index in range(amdsmi_interface.AMDSMI_MAX_NUM_CLKS):
vclk_index = f"vclk_{clock_index}"
clocks[vclk_index] = {"clk" : "N/A",
"min_clk" : "N/A",
"max_clk" : "N/A",
"clk_locked" : "N/A",
"deep_sleep" : "N/A"}
for clock_index in range(amdsmi_interface.AMDSMI_MAX_NUM_CLKS):
dclk_index = f"dclk_{clock_index}"
clocks[dclk_index] = {"clk" : "N/A",
"min_clk" : "N/A",
"max_clk" : "N/A",
"clk_locked" : "N/A",
"deep_sleep" : "N/A"}
clocks["fclk_0"] = {"clk" : "N/A",
"min_clk" : "N/A",
"max_clk" : "N/A",
"clk_locked" : "N/A",
"deep_sleep" : "N/A"}
clocks["socclk_0"] = {"clk" : "N/A",
"min_clk" : "N/A",
"max_clk" : "N/A",
"clk_locked" : "N/A",
"deep_sleep" : "N/A"}
clock_unit = "MHz"
# Populate clock values from gpu_metrics_info
# Populate GFX clock values
try:
current_gfx_clocks = gpu_metric["current_gfxclks"]
if current_gfx_clocks != "N/A":
for clock_index, current_gfx_clock in enumerate(current_gfx_clocks):
# If the current clock is N/A then nothing else applies
if current_gfx_clock == "N/A":
continue
gfx_index = f"gfx_{clock_index}"
clocks[gfx_index]["clk"] = self.helpers.unit_format(self.logger,
current_gfx_clock,
clock_unit)
# Populate clock locked status
if gpu_metric["gfxclk_lock_status"] != "N/A":
gfx_clock_lock_flag = 1 << clock_index # This is the position of the clock lock flag
if gpu_metric["gfxclk_lock_status"] & gfx_clock_lock_flag:
clocks[gfx_index]["clk_locked"] = "ENABLED"
else:
clocks[gfx_index]["clk_locked"] = "DISABLED"
except Exception as e:
logging.debug("Failed to get current_gfxclks for gpu %s | %s", gpu_id, e)
# Populate MEM clock value
try:
current_mem_clock = gpu_metric["current_uclk"] # single value
if current_mem_clock != "N/A":
clocks["mem_0"]["clk"] = self.helpers.unit_format(self.logger,
current_mem_clock,
clock_unit)
except Exception as e:
logging.debug("Failed to get current_uclk for gpu %s | %s", gpu_id, e)
# Populate VCLK clock values
try:
current_vclk_clocks = gpu_metric["current_vclk0s"]
# If the current vclk clocks are not available, we cannot proceed further
if current_vclk_clocks != "N/A":
for clock_index, current_vclk_clock in enumerate(current_vclk_clocks):
# If the current clock is N/A then nothing else applies
if current_vclk_clock == "N/A":
continue
vclk_index = f"vclk_{clock_index}"
clocks[vclk_index]["clk"] = self.helpers.unit_format(self.logger,
current_vclk_clock,
clock_unit)
except Exception as e:
logging.debug("Failed to get current_vclk0s for gpu %s | %s", gpu_id, e)
# Populate DCLK clock values
try:
current_dclk_clocks = gpu_metric["current_dclk0s"]
# If the current dclk clocks are not available, we cannot proceed further
if current_dclk_clocks != "N/A":
for clock_index, current_dclk_clock in enumerate(current_dclk_clocks):
# If the current clock is N/A then nothing else applies
if current_dclk_clock == "N/A":
continue
dclk_index = f"dclk_{clock_index}"
clocks[dclk_index]["clk"] = self.helpers.unit_format(self.logger,
current_dclk_clock,
clock_unit)
except Exception as e:
logging.debug("Failed to get current_dclk0s for gpu %s | %s", gpu_id, e)
# Populate FCLK clock value; fclk not present in gpu_metrics so use amdsmi_get_clk_freq
try:
frequency_dict = amdsmi_interface.amdsmi_get_clk_freq(args.gpu, amdsmi_interface.AmdSmiClkType.DF)
current_fclk_clock = frequency_dict['frequency'][frequency_dict['current']]
current_fclk_clock = self.helpers.convert_SI_unit(current_fclk_clock, self.helpers.SI_Unit.MICRO)
clocks["fclk_0"]["clk"] = self.helpers.unit_format(self.logger,
current_fclk_clock,
clock_unit)
except (KeyError, amdsmi_exception.AmdSmiLibraryException) as e:
logging.debug("Failed to get fclk info for gpu %s | %s", gpu_id, e)
# Populate SOCCLK clock value
try:
current_socclk_clock = gpu_metric["current_socclk"]
# If the current socclk clocks are not available, we cannot proceed further
if current_socclk_clock != "N/A":
clocks["socclk_0"]["clk"] = self.helpers.unit_format(self.logger,
current_socclk_clock,
clock_unit)
except KeyError as e:
logging.debug("Failed to get current_socclk for gpu %s | %s", gpu_id, e)
# Populate the max and min clock values from sysfs.
# Min and Max values are per clock type, not per clock engine.
# Populate the deep sleep value from amdsmi_get_clock_info
# GFX min and max clocks
try:
gfx_clock_info_dict = amdsmi_interface.amdsmi_get_clock_info(args.gpu,
amdsmi_interface.AmdSmiClkType.GFX)
for clock_index in range(amdsmi_interface.AMDSMI_MAX_NUM_GFX_CLKS):
gfx_index = f"gfx_{clock_index}"
if clocks[gfx_index]["clk"] == "N/A":
# if the current clock is N/A then we shouldn't populate the max and min values
continue
clocks[gfx_index]["min_clk"] = self.helpers.unit_format(self.logger,
gfx_clock_info_dict["min_clk"],
clock_unit)
clocks[gfx_index]["max_clk"] = self.helpers.unit_format(self.logger,
gfx_clock_info_dict["max_clk"],
clock_unit)
# Add the clk_deep_sleep
clocks[gfx_index]["deep_sleep"] = gfx_clock_info_dict["clk_deep_sleep"]
except (KeyError, amdsmi_exception.AmdSmiLibraryException) as e:
logging.debug("Failed to get gfx clock info for gpu %s | %s", gpu_id, e)
# MEM min and max clocks
try:
mem_clock_info_dict = amdsmi_interface.amdsmi_get_clock_info(args.gpu,
amdsmi_interface.AmdSmiClkType.MEM)
# if the current clock is N/A then we shouldn't populate the max and min values
if clocks["mem_0"]["clk"] != "N/A":
clocks["mem_0"]["min_clk"] = self.helpers.unit_format(self.logger,
mem_clock_info_dict["min_clk"],
clock_unit)
clocks["mem_0"]["max_clk"] = self.helpers.unit_format(self.logger,
mem_clock_info_dict["max_clk"],
clock_unit)
# Add the clk_deep_sleep
clocks["mem_0"]["deep_sleep"] = mem_clock_info_dict["clk_deep_sleep"]
except (KeyError, amdsmi_exception.AmdSmiLibraryException) as e:
logging.debug("Failed to get mem clock info for gpu %s | %s", gpu_id, e)
# VCLK min and max clocks
try:
# Retrieve clock information for VCLK0 (Video Clock 0)
vclk_clock_info_dict = amdsmi_interface.amdsmi_get_clock_info(args.gpu, amdsmi_interface.AmdSmiClkType.VCLK0)
# Iterate through the maximum number of VCLK clocks supported
for index in range(amdsmi_interface.AMDSMI_MAX_NUM_CLKS):
vclk_index = f"vclk_{index}" # Construct the index key for the clock
# Check if the current clock value is not "N/A"
if clocks[vclk_index]["clk"] != "N/A":
# Format and assign the minimum clock value for the current VCLK
clocks[vclk_index]["min_clk"] = self.helpers.unit_format(self.logger,
vclk_clock_info_dict["min_clk"],
clock_unit)
# Format and assign the maximum clock value for the current VCLK
clocks[vclk_index]["max_clk"] = self.helpers.unit_format(self.logger,
vclk_clock_info_dict["max_clk"],
clock_unit)
# Add the clk_deep_sleep
clocks[vclk_index]["deep_sleep"] = vclk_clock_info_dict["clk_deep_sleep"]
except (KeyError, amdsmi_exception.AmdSmiLibraryException) as e:
# Log a debug message if retrieving VCLK clock information fails
logging.debug("Failed to get vclk clock info for gpu %s | %s", gpu_id, e)
# DCLK min and max clocks
try:
# Retrieve clock information for DCLK0 (Display Clock 0)
dclk_clock_info_dict = amdsmi_interface.amdsmi_get_clock_info(args.gpu, amdsmi_interface.AmdSmiClkType.DCLK0)
# Iterate through the maximum number of DCLK clocks supported
for index in range(amdsmi_interface.AMDSMI_MAX_NUM_CLKS):
dclk_index = f"dclk_{index}" # Construct the index key for the clock
# Check if the current clock value is not "N/A"
if clocks[dclk_index]["clk"] != "N/A":
# Format and assign the minimum clock value for the current DCLK
clocks[dclk_index]["min_clk"] = self.helpers.unit_format(self.logger,
dclk_clock_info_dict["min_clk"],
clock_unit)
# Format and assign the maximum clock value for the current DCLK
clocks[dclk_index]["max_clk"] = self.helpers.unit_format(self.logger,
dclk_clock_info_dict["max_clk"],
clock_unit)
# Add the clk_deep_sleep
clocks[dclk_index]["deep_sleep"] = dclk_clock_info_dict["clk_deep_sleep"]
except (KeyError, amdsmi_exception.AmdSmiLibraryException) as e:
logging.debug("Failed to get dclk clock info for gpu %s | %s", gpu_id, e)
# FCLK min and max clocks
try:
fclk_clk_info_dict = amdsmi_interface.amdsmi_get_clock_info(args.gpu,
amdsmi_interface.AmdSmiClkType.DF)
# if the current clock is N/A then we shouldn't populate the max and min values
if clocks["fclk_0"]["clk"] != "N/A":
clocks["fclk_0"]["min_clk"] = self.helpers.unit_format(self.logger,
fclk_clk_info_dict["min_clk"],
clock_unit)
clocks["fclk_0"]["max_clk"] = self.helpers.unit_format(self.logger,
fclk_clk_info_dict["max_clk"],
clock_unit)
# Add the clk_deep_sleep
clocks["fclk_0"]["deep_sleep"] = fclk_clk_info_dict["clk_deep_sleep"]
except amdsmi_exception.AmdSmiLibraryException as e:
logging.debug("Failed to get fclk info for gpu %s | %s", gpu_id, e.get_error_info())
# SOCCLK min and max clocks
try:
socclk_clk_info_dict = amdsmi_interface.amdsmi_get_clock_info(args.gpu,
amdsmi_interface.AmdSmiClkType.SOC)
# if the current clock is N/A then we shouldn't populate the max and min values
if clocks["socclk_0"]["clk"] != "N/A":
clocks["socclk_0"]["min_clk"] = self.helpers.unit_format(self.logger,
socclk_clk_info_dict["min_clk"],
clock_unit)
clocks["socclk_0"]["max_clk"] = self.helpers.unit_format(self.logger,
socclk_clk_info_dict["max_clk"],
clock_unit)
# Add the clk_deep_sleep
clocks["socclk_0"]["deep_sleep"] = socclk_clk_info_dict["clk_deep_sleep"]
except amdsmi_exception.AmdSmiLibraryException as e:
logging.debug("Failed to get socclk info for gpu %s | %s", gpu_id, e.get_error_info())
# Iterate over each clock and its data to determine if deep sleep is enabled
# based on the comparison between the current clock value and the minimum clock value.
for clock, clock_data in clocks.items():
clk_value = 0
min_clk_value = 0
try:
clk = clock_data["clk"]
min_clk = clock_data["min_clk"]
if clk == "N/A" or min_clk == "N/A":
continue
# Extract numeric value if clk/min_clk is a dict, else use as is
if isinstance(clk, dict):
clk_value = int(clk.get("value", 0))
else:
if isinstance(clk, str):
clk_value = int(str(clk).split()[0])
else:
clk_value = int(clk)
if isinstance(min_clk, dict):
min_clk_value = int(min_clk.get("value", 0))
else:
if isinstance(min_clk, str):
min_clk_value = int(str(min_clk).split()[0])
else:
min_clk_value = int(min_clk)
# If the clk value is less than the min_clk value, then deep sleep is enabled
if clk_value < min_clk_value:
clock_data["deep_sleep"] = "ENABLED"
else:
clock_data["deep_sleep"] = "DISABLED"
except Exception as e:
logging.debug("Failed to get deep sleep status for gpu %s | %s", gpu_id, e)
values_dict['clock'] = clocks
if "temperature" in current_platform_args:
if args.temperature:
try:
temperature_edge_current = amdsmi_interface.amdsmi_get_temp_metric(
args.gpu, amdsmi_interface.AmdSmiTemperatureType.EDGE, amdsmi_interface.AmdSmiTemperatureMetric.CURRENT)
except amdsmi_exception.AmdSmiLibraryException as e:
temperature_edge_current = "N/A"
logging.debug("Failed to get current edge temperature for gpu %s | %s", gpu_id, e.get_error_info())
try:
temperature_edge_limit = amdsmi_interface.amdsmi_get_temp_metric(
args.gpu, amdsmi_interface.AmdSmiTemperatureType.EDGE, amdsmi_interface.AmdSmiTemperatureMetric.CRITICAL)
except amdsmi_exception.AmdSmiLibraryException as e:
temperature_edge_limit = "N/A"
logging.debug("Failed to get edge temperature limit for gpu %s | %s", gpu_id, e.get_error_info())
# If edge limit is reporting 0 then set the current edge temp to N/A
if temperature_edge_limit == 0:
temperature_edge_current = "N/A"
try:
temperature_hotspot_current = amdsmi_interface.amdsmi_get_temp_metric(
args.gpu, amdsmi_interface.AmdSmiTemperatureType.HOTSPOT, amdsmi_interface.AmdSmiTemperatureMetric.CURRENT)
except amdsmi_exception.AmdSmiLibraryException as e:
temperature_hotspot_current = "N/A"
logging.debug("Failed to get current hotspot temperature for gpu %s | %s", gpu_id, e.get_error_info())
try:
temperature_vram_current = amdsmi_interface.amdsmi_get_temp_metric(
args.gpu, amdsmi_interface.AmdSmiTemperatureType.VRAM, amdsmi_interface.AmdSmiTemperatureMetric.CURRENT)
except amdsmi_exception.AmdSmiLibraryException as e:
temperature_vram_current = "N/A"
logging.debug("Failed to get current vram temperature for gpu %s | %s", gpu_id, e.get_error_info())
temperatures = {'edge': temperature_edge_current,
'hotspot': temperature_hotspot_current,
'mem': temperature_vram_current}
temp_unit_human_readable = '\N{DEGREE SIGN}C'
temp_unit_json = 'C'
for temperature_key, temperature_value in temperatures.items():
if 'N/A' not in str(temperature_value):
if self.logger.is_human_readable_format():
temperatures[temperature_key] = f"{temperature_value} {temp_unit_human_readable}"
if self.logger.is_json_format():
temperatures[temperature_key] = {"value" : temperature_value,
"unit" : temp_unit_json}
values_dict['temperature'] = temperatures
# Since pcie bw may increase based on frequent metrics calls, we add it to the output here, but the populate the values first
if "pcie" in current_platform_args:
if args.pcie:
values_dict['pcie'] = pcie_dict
if "gpu_board" in current_platform_args:
if args.gpu_board:
gpu_board_temp_dict = self.helpers.get_gpu_board_temperatures(args.gpu, gpu_id, self.logger)
# if every value is N/A, then we don't want to display the values unless explicitly told to
# all args_list being True indicates that this gpu_board is not explicitly called itself
args_list = [getattr(args, arg) for arg in current_platform_args]
if all(value == "N/A" for value in gpu_board_temp_dict.values()) and all(arg == True for arg in args_list):
gpu_board_temp_dict = {}
if gpu_board_temp_dict:
values_dict['gpu_board'] = {'temperature':gpu_board_temp_dict}
if "base_board" in current_platform_args:
if args.base_board:
base_board_temp_dict = self.helpers.get_base_board_temperatures(args.gpu, gpu_id, self.logger)
# if every value is N/A, then we don't want to display the values unless explicitly told to
# all args_list being True indicates that this base_board is not explicitly called itself
args_list = [getattr(args, arg) for arg in current_platform_args]
if all(value == "N/A" for value in base_board_temp_dict.values()) and all(arg == True for arg in args_list):
base_board_temp_dict = {}
if base_board_temp_dict:
values_dict['base_board'] = {'temperature':base_board_temp_dict}
if "ecc" in current_platform_args:
if args.ecc:
ecc_count = {}
try:
ecc_count = amdsmi_interface.amdsmi_get_gpu_total_ecc_count(args.gpu)
ecc_count['total_correctable_count'] = ecc_count.pop('correctable_count')
ecc_count['total_uncorrectable_count'] = ecc_count.pop('uncorrectable_count')
ecc_count['total_deferred_count'] = ecc_count.pop('deferred_count')
except amdsmi_exception.AmdSmiLibraryException as e:
ecc_count['total_correctable_count'] = "N/A"
ecc_count['total_uncorrectable_count'] = "N/A"
ecc_count['cache_correctable_count'] = "N/A"
ecc_count['cache_uncorrectable_count'] = "N/A"
logging.debug("Failed to get total ecc count for gpu %s | %s", gpu_id, e.get_error_info())
if ecc_count['total_correctable_count'] != "N/A":
# Get the UMC error count for getting total cache correctable errors
umc_block = amdsmi_interface.AmdSmiGpuBlock['UMC']
try:
umc_count = amdsmi_interface.amdsmi_get_gpu_ecc_count(args.gpu, umc_block)
ecc_count['cache_correctable_count'] = ecc_count['total_correctable_count'] - umc_count['correctable_count']
ecc_count['cache_uncorrectable_count'] = ecc_count['total_uncorrectable_count'] - umc_count['uncorrectable_count']
except amdsmi_exception.AmdSmiLibraryException as e:
ecc_count['cache_correctable_count'] = "N/A"
ecc_count['cache_uncorrectable_count'] = "N/A"
logging.debug("Failed to get cache ecc count for gpu %s at block %s | %s", gpu_id, umc_block, e.get_error_info())
values_dict['ecc'] = ecc_count
if "ecc_blocks" in current_platform_args:
if args.ecc_blocks:
ecc_dict = {}
sysfs_blocks = ["UMC", "SDMA", "GFX", "MMHUB", "PCIE_BIF", "HDP", "XGMI_WAFL"]
try:
ras_states = amdsmi_interface.amdsmi_get_gpu_ras_block_features_enabled(args.gpu)
for state in ras_states:
# Only add enabled blocks that are also in sysfs
if state['status'] == amdsmi_interface.AmdSmiRasErrState.ENABLED.name:
gpu_block = amdsmi_interface.AmdSmiGpuBlock[state['block']]
# if the blocks are uncountable do not add them at all.
if gpu_block.name in sysfs_blocks:
try:
ecc_count = amdsmi_interface.amdsmi_get_gpu_ecc_count(args.gpu, gpu_block)
ecc_dict[state['block']] = {'correctable_count' : ecc_count['correctable_count'],
'uncorrectable_count' : ecc_count['uncorrectable_count'],
'deferred_count' : ecc_count['deferred_count']}
except amdsmi_exception.AmdSmiLibraryException as e:
ecc_dict[state['block']] = {'correctable_count' : "N/A",
'uncorrectable_count' : "N/A",
'deferred_count' : "N/A"}
logging.debug("Failed to get ecc count for gpu %s at block %s | %s", gpu_id, gpu_block, e.get_error_info())
values_dict['ecc_blocks'] = ecc_dict
except amdsmi_exception.AmdSmiLibraryException as e:
values_dict['ecc_blocks'] = "N/A"
logging.debug("Failed to get ecc block features for gpu %s | %s", gpu_id, e.get_error_info())
if "fan" in current_platform_args:
if args.fan:
fan_dict = {"speed" : "N/A",
"max" : "N/A",
"rpm" : "N/A",
"usage" : "N/A"}
try:
fan_speed = amdsmi_interface.amdsmi_get_gpu_fan_speed(args.gpu, 0)
fan_dict["speed"] = fan_speed
except amdsmi_exception.AmdSmiLibraryException as e:
logging.debug("Failed to get fan speed for gpu %s | %s", args.gpu, e.get_error_info())
try:
fan_max = amdsmi_interface.amdsmi_get_gpu_fan_speed_max(args.gpu, 0)
fan_usage = "N/A"
if fan_max > 0 and fan_dict["speed"] != "N/A":
fan_usage = round((float(fan_speed) / float(fan_max)) * 100, 2)
fan_usage_unit = '%'
if self.logger.is_human_readable_format():
fan_usage = f"{fan_usage} {fan_usage_unit}"
if self.logger.is_json_format():
fan_usage = {"value" : fan_usage,
"unit" : fan_usage_unit}
fan_dict["max"] = fan_max
fan_dict["usage"] = fan_usage
except amdsmi_exception.AmdSmiLibraryException as e:
logging.debug("Failed to get fan max speed for gpu %s | %s", args.gpu, e.get_error_info())
try:
fan_rpm = amdsmi_interface.amdsmi_get_gpu_fan_rpms(args.gpu, 0)
fan_dict["rpm"] = fan_rpm
except amdsmi_exception.AmdSmiLibraryException as e:
logging.debug("Failed to get fan rpms for gpu %s | %s", args.gpu, e.get_error_info())
values_dict["fan"] = fan_dict
if "voltage_curve" in current_platform_args:
if args.voltage_curve:
# Populate N/A values per voltage point
voltage_point_dict = {}
for point in range(amdsmi_interface.AMDSMI_NUM_VOLTAGE_CURVE_POINTS):
voltage_point_dict[f'point_{point}_frequency'] = "N/A"
voltage_point_dict[f'point_{point}_voltage'] = "N/A"
try:
od_volt = amdsmi_interface.amdsmi_get_gpu_od_volt_info(args.gpu)
logging.debug(f"OD Voltage info: {od_volt}")
except amdsmi_exception.AmdSmiLibraryException as e:
od_volt = "N/A" # Value not used, but needs to not be a dict
logging.debug("Failed to get voltage curve for gpu %s | %s", gpu_id, e.get_error_info())
# Populate voltage point values
for point in range(amdsmi_interface.AMDSMI_NUM_VOLTAGE_CURVE_POINTS):
if isinstance(od_volt, dict):
logging.debug(f"point_{point} frequency: {od_volt['curve.vc_points'][point]['frequency']}")
logging.debug(f"point_{point} voltage: {od_volt['curve.vc_points'][point]['voltage']}")
frequency = int(od_volt["curve.vc_points"][point]['frequency'] / 1000000)
voltage = int(od_volt["curve.vc_points"][point]['voltage'])
else:
frequency = "N/A"
voltage = "N/A"
if frequency == 0:
frequency = "N/A"
if voltage == 0:
voltage = "N/A"
if frequency != "N/A":
frequency = self.helpers.unit_format(self.logger, frequency, "Mhz")
if voltage != "N/A":
voltage = self.helpers.unit_format(self.logger, voltage, "mV")
voltage_point_dict[f'point_{point}_frequency'] = frequency
voltage_point_dict[f'point_{point}_voltage'] = voltage
values_dict['voltage_curve'] = voltage_point_dict
if "overdrive" in current_platform_args:
if args.overdrive:
try:
overdrive_level = amdsmi_interface.amdsmi_get_gpu_overdrive_level(args.gpu)
od_unit = '%'
values_dict['overdrive'] = self.helpers.unit_format(self.logger, overdrive_level, od_unit)
except amdsmi_exception.AmdSmiLibraryException as e:
values_dict['overdrive'] = "N/A"
logging.debug("Failed to get gpu overdrive level for gpu %s | %s", gpu_id, e.get_error_info())
try:
mem_overdrive_level = amdsmi_interface.amdsmi_get_gpu_mem_overdrive_level(args.gpu)
od_unit = '%'
values_dict['mem_overdrive'] = self.helpers.unit_format(self.logger, mem_overdrive_level, od_unit)
except amdsmi_exception.AmdSmiLibraryException as e:
values_dict['mem_overdrive'] = "N/A"
logging.debug("Failed to get mem overdrive level for gpu %s | %s", gpu_id, e.get_error_info())
if "perf_level" in current_platform_args:
if args.perf_level:
try:
perf_level = amdsmi_interface.amdsmi_get_gpu_perf_level(args.gpu)
values_dict['perf_level'] = perf_level
except amdsmi_exception.AmdSmiLibraryException as e:
values_dict['perf_level'] = "N/A"
logging.debug("Failed to get perf level for gpu %s | %s", gpu_id, e.get_error_info())
if "xgmi_err" in current_platform_args:
if args.xgmi_err:
try:
xgmi_err_status = amdsmi_interface.amdsmi_gpu_xgmi_error_status(args.gpu)
values_dict['xgmi_err'] = amdsmi_interface.amdsmi_wrapper.amdsmi_xgmi_status_t__enumvalues[xgmi_err_status]
except amdsmi_exception.AmdSmiLibraryException as e:
values_dict['xgmi_err'] = "N/A"
logging.debug("Failed to get xgmi error status for gpu %s | %s", gpu_id, e.get_error_info())
if "voltage" in current_platform_args:
if args.voltage:
voltage_dict = {}
all_voltage = {
"vddboard": amdsmi_interface.AmdSmiVoltageType.VDDBOARD
}
for volt_type, volt_metric in all_voltage.items():
try:
voltage = amdsmi_interface.amdsmi_get_gpu_volt_metric(args.gpu, volt_metric, amdsmi_interface.AmdSmiVoltageMetric.CURRENT)
if voltage == 0:
voltage = "N/A"
voltage_dict[volt_type] = self.helpers.unit_format(self.logger, voltage, "mV")
except amdsmi_exception.AmdSmiLibraryException as e:
voltage_dict[volt_type] = "N/A"
logging.debug("Failed to get voltage for gpu %s | %s", gpu_id, e.get_error_info())
values_dict['voltage'] = voltage_dict
if "energy" in current_platform_args:
if args.energy:
try:
energy_dict = amdsmi_interface.amdsmi_get_energy_count(args.gpu)
energy = round(energy_dict["energy_accumulator"] * energy_dict["counter_resolution"], 3)
energy /= 1000000
energy = round(energy, 3)
energy_unit = 'J'
if self.logger.is_human_readable_format():
energy = f"{energy} {energy_unit}"
if self.logger.is_json_format():
energy = {"value" : energy,
"unit" : energy_unit}
values_dict['energy'] = {"total_energy_consumption" : energy}
except amdsmi_interface.AmdSmiLibraryException as e:
values_dict['energy'] = "N/A"
logging.debug("Failed to get energy usage for gpu %s | %s", args.gpu, e.get_error_info())
if "mem_usage" in current_platform_args:
if args.mem_usage:
memory_usage = {'total_vram': "N/A",
'used_vram': "N/A",
'free_vram': "N/A",
'total_visible_vram': "N/A",
'used_visible_vram': "N/A",
'free_visible_vram': "N/A",
'total_gtt': "N/A",
'used_gtt': "N/A",
'free_gtt': "N/A"}
# Total VRAM
try:
total_vram = amdsmi_interface.amdsmi_get_gpu_memory_total(args.gpu, amdsmi_interface.AmdSmiMemoryType.VRAM)
memory_usage['total_vram'] = total_vram // (1024*1024)
except amdsmi_exception.AmdSmiLibraryException as e:
logging.debug("Failed to get total VRAM memory for gpu %s | %s", gpu_id, e.get_error_info())
try:
total_visible_vram = amdsmi_interface.amdsmi_get_gpu_memory_total(args.gpu, amdsmi_interface.AmdSmiMemoryType.VIS_VRAM)
memory_usage['total_visible_vram'] = total_visible_vram // (1024*1024)
except amdsmi_exception.AmdSmiLibraryException as e:
logging.debug("Failed to get total VIS VRAM memory for gpu %s | %s", gpu_id, e.get_error_info())
try:
total_gtt = amdsmi_interface.amdsmi_get_gpu_memory_total(args.gpu, amdsmi_interface.AmdSmiMemoryType.GTT)
memory_usage['total_gtt'] = total_gtt // (1024*1024)
except amdsmi_exception.AmdSmiLibraryException as e:
logging.debug("Failed to get total GTT memory for gpu %s | %s", gpu_id, e.get_error_info())
# Used VRAM
try:
used_vram = amdsmi_interface.amdsmi_get_gpu_memory_usage(args.gpu, amdsmi_interface.AmdSmiMemoryType.VRAM)
memory_usage['used_vram'] = used_vram // (1024*1024)
except amdsmi_exception.AmdSmiLibraryException as e:
logging.debug("Failed to get used VRAM memory for gpu %s | %s", gpu_id, e.get_error_info())
try:
used_visible_vram = amdsmi_interface.amdsmi_get_gpu_memory_usage(args.gpu, amdsmi_interface.AmdSmiMemoryType.VIS_VRAM)
memory_usage['used_visible_vram'] = used_visible_vram // (1024*1024)
except amdsmi_exception.AmdSmiLibraryException as e:
logging.debug("Failed to get used VIS VRAM memory for gpu %s | %s", gpu_id, e.get_error_info())
try:
used_gtt = amdsmi_interface.amdsmi_get_gpu_memory_usage(args.gpu, amdsmi_interface.AmdSmiMemoryType.GTT)
memory_usage['used_gtt'] = used_gtt // (1024*1024)
except amdsmi_exception.AmdSmiLibraryException as e:
logging.debug("Failed to get used GTT memory for gpu %s | %s", gpu_id, e.get_error_info())
# Free VRAM
if memory_usage['total_vram'] != "N/A" and memory_usage['used_vram'] != "N/A":
memory_usage['free_vram'] = memory_usage['total_vram'] - memory_usage['used_vram']
if memory_usage['total_visible_vram'] != "N/A" and memory_usage['used_visible_vram'] != "N/A":
memory_usage['free_visible_vram'] = memory_usage['total_visible_vram'] - memory_usage['used_visible_vram']
if memory_usage['total_gtt'] != "N/A" and memory_usage['used_gtt'] != "N/A":
memory_usage['free_gtt'] = memory_usage['total_gtt'] - memory_usage['used_gtt']
memory_unit = 'MB'
for key, value in memory_usage.items():
if value != "N/A":
if self.logger.is_human_readable_format():
memory_usage[key] = f"{value} {memory_unit}"
if self.logger.is_json_format():
memory_usage[key] = {"value" : value,
"unit" : memory_unit}
values_dict['mem_usage'] = memory_usage
if "throttle" in current_platform_args:
if args.throttle:
throttle_status = {
# Current values - counter/accumulated
'accumulation_counter': "N/A",
'prochot_accumulated': "N/A",
'ppt_accumulated': "N/A",
'socket_thermal_accumulated': "N/A",
'vr_thermal_accumulated': "N/A",
'hbm_thermal_accumulated': "N/A",
'gfx_clk_below_host_limit_accumulated': "N/A", # deprecated
'gfx_clk_below_host_limit_power_accumulated': "N/A",
'gfx_clk_below_host_limit_thermal_accumulated': "N/A",
'total_gfx_clk_below_host_limit_accumulated': "N/A",
'low_utilization_accumulated': "N/A",
# violation status values - active/not active
'prochot_violation_status': "N/A",
'ppt_violation_status': "N/A",
'socket_thermal_violation_status': "N/A",
'vr_thermal_violation_status': "N/A",
'hbm_thermal_violation_status': "N/A",
'gfx_clk_below_host_limit_violation_status': "N/A", # deprecated
'gfx_clk_below_host_limit_power_violation_status': "N/A",
'gfx_clk_below_host_limit_thermal_violation_status': "N/A",
'total_gfx_clk_below_host_limit_violation_status': "N/A",
'low_utilization_violation_status': "N/A",
# violation activity values - percent
'prochot_violation_activity': "N/A",
'ppt_violation_activity': "N/A",
'socket_thermal_violation_activity': "N/A",
'vr_thermal_violation_activity': "N/A",
'hbm_thermal_violation_activity': "N/A",
'gfx_clk_below_host_limit_violation_activity': "N/A", # deprecated
'gfx_clk_below_host_limit_power_violation_activity': "N/A",
'gfx_clk_below_host_limit_thermal_violation_activity': "N/A",
'total_gfx_clk_below_host_limit_violation_activity': "N/A",
'low_utilization_violation_activity': "N/A",
}
try:
violation_status = amdsmi_interface.amdsmi_get_violation_status(args.gpu)
throttle_status['accumulation_counter'] = violation_status['acc_counter']
throttle_status['prochot_accumulated'] = violation_status['acc_prochot_thrm']
throttle_status['ppt_accumulated'] = violation_status['acc_ppt_pwr']
throttle_status['socket_thermal_accumulated'] = violation_status['acc_socket_thrm']
throttle_status['vr_thermal_accumulated'] = violation_status['acc_vr_thrm']
throttle_status['hbm_thermal_accumulated'] = violation_status['acc_hbm_thrm']
throttle_status['gfx_clk_below_host_limit_accumulated'] = violation_status['acc_gfx_clk_below_host_limit'] #deprecated
throttle_status['gfx_clk_below_host_limit_power_accumulated'] = self.helpers.build_xcp_dict('acc_gfx_clk_below_host_limit_pwr', violation_status, num_partition)
throttle_status['gfx_clk_below_host_limit_thermal_accumulated'] = self.helpers.build_xcp_dict('acc_gfx_clk_below_host_limit_thm', violation_status, num_partition)
throttle_status['total_gfx_clk_below_host_limit_accumulated'] = self.helpers.build_xcp_dict('acc_gfx_clk_below_host_limit_total', violation_status, num_partition)
throttle_status['low_utilization_accumulated'] = self.helpers.build_xcp_dict('acc_low_utilization', violation_status, num_partition)
throttle_status['prochot_violation_status'] = self.helpers.build_xcp_dict('active_prochot_thrm', violation_status, num_partition)
throttle_status['ppt_violation_status'] = self.helpers.build_xcp_dict('active_ppt_pwr', violation_status, num_partition)
throttle_status['socket_thermal_violation_status'] = self.helpers.build_xcp_dict('active_socket_thrm', violation_status, num_partition)
throttle_status['vr_thermal_violation_status'] = self.helpers.build_xcp_dict('active_vr_thrm', violation_status, num_partition)
throttle_status['hbm_thermal_violation_status'] = self.helpers.build_xcp_dict('active_hbm_thrm', violation_status, num_partition)
throttle_status['gfx_clk_below_host_limit_violation_status'] = self.helpers.build_xcp_dict('active_gfx_clk_below_host_limit', violation_status, num_partition) # deprecated
throttle_status['gfx_clk_below_host_limit_power_violation_status'] = self.helpers.build_xcp_dict('active_gfx_clk_below_host_limit_pwr', violation_status, num_partition)
throttle_status['gfx_clk_below_host_limit_thermal_violation_status'] = self.helpers.build_xcp_dict('active_gfx_clk_below_host_limit_thm', violation_status, num_partition)
throttle_status['total_gfx_clk_below_host_limit_violation_status'] = self.helpers.build_xcp_dict('active_gfx_clk_below_host_limit_total', violation_status, num_partition)
throttle_status['low_utilization_violation_status'] = self.helpers.build_xcp_dict('active_low_utilization', violation_status, num_partition)
throttle_status['prochot_violation_activity'] = violation_status['per_prochot_thrm']
throttle_status['ppt_violation_activity'] = violation_status['per_ppt_pwr']
throttle_status['socket_thermal_violation_activity'] = violation_status['per_socket_thrm']
throttle_status['vr_thermal_violation_activity'] = violation_status['per_vr_thrm']
throttle_status['hbm_thermal_violation_activity'] = violation_status['per_hbm_thrm']
throttle_status['gfx_clk_below_host_limit_violation_activity'] = violation_status['per_gfx_clk_below_host_limit'] # deprecated
throttle_status['gfx_clk_below_host_limit_power_violation_activity'] = self.helpers.build_xcp_dict('per_gfx_clk_below_host_limit_pwr', violation_status, num_partition)
throttle_status['gfx_clk_below_host_limit_thermal_violation_activity'] = self.helpers.build_xcp_dict('per_gfx_clk_below_host_limit_thm', violation_status, num_partition)
throttle_status['total_gfx_clk_below_host_limit_violation_activity'] = self.helpers.build_xcp_dict('per_gfx_clk_below_host_limit_total', violation_status, num_partition)
throttle_status['low_utilization_violation_activity'] = self.helpers.build_xcp_dict('per_low_utilization', violation_status, num_partition)
except amdsmi_exception.AmdSmiLibraryException as e:
values_dict['throttle'] = throttle_status
logging.debug("Failed to get violation status' for gpu %s | %s", gpu_id, e.get_error_info())
for key, value in throttle_status.items():
activity_unit = ''
if "_activity" in key:
activity_unit = '%'
if self.logger.is_human_readable_format():
if isinstance(value, (list, dict)):
for k, v in value.items():
for index, activity in enumerate(v):
value[k][index] = self.helpers.unit_format(self.logger, activity, activity_unit)
value[k] = '[' + ", ".join(value[k]) + ']'
elif value != "N/A":
throttle_status[key] = self.helpers.unit_format(self.logger, value, activity_unit)
if self.logger.is_json_format():
if isinstance(value, (list, dict)):
for k, v in value.items():
for index, activity in enumerate(v):
value[k][index] = self.helpers.unit_format(self.logger, activity, activity_unit)
elif value != "N/A":
throttle_status[key] = self.helpers.unit_format(self.logger, value, activity_unit)
values_dict['throttle'] = throttle_status
# Store timestamp first if watching_output is enabled
if watching_output:
self.logger.store_output(args.gpu, 'timestamp', int(time.time()))
self.logger.store_output(args.gpu, 'values', values_dict)
self.logger.store_gpu_json_output.append(values_dict)
if multiple_devices:
self.logger.store_multiple_device_output()
return # Skip printing when there are multiple devices
if not self.logger.is_json_format():
self.logger.print_output(watching_output=watching_output)
if watching_output: # End of single gpu add to watch_output
self.logger.store_watch_output(multiple_device_enabled=False)
def metric_cpu(self, args, multiple_devices=False, cpu=None, cpu_power_metrics=None, cpu_prochot=None,
cpu_freq_metrics=None, cpu_c0_res=None, cpu_lclk_dpm_level=None,
cpu_pwr_svi_telemetry_rails=None, cpu_io_bandwidth=None, cpu_xgmi_bandwidth=None,
cpu_metrics_ver=None, cpu_metrics_table=None, cpu_socket_energy=None,
cpu_ddr_bandwidth=None, cpu_temp=None, cpu_dimm_temp_range_rate=None,
cpu_dimm_pow_consumption=None, cpu_dimm_thermal_sensor=None,
cpu_dfcstate_ctrl=None, cpu_railisofreq_policy=None):
"""Get Metric information for target cpu
Args:
args (Namespace): Namespace containing the parsed CLI args
multiple_devices (bool, optional): True if checking for multiple devices. Defaults to False.
cpu (cpu_handle, optional): device_handle for target device. Defaults to None.
cpu_power_metrics (bool, optional): Value override for args.cpu_power_metrics. Defaults to None
cpu_prochot (bool, optional): Value override for args.cpu_prochot. Defaults to None.
cpu_freq_metrics (bool, optional): Value override for args.cpu_freq_metrics. Defaults to None.
cpu_c0_res (bool, optional): Value override for args.cpu_c0_res. Defaults to None
cpu_lclk_dpm_level (list, optional): Value override for args.cpu_lclk_dpm_level. Defaults to None
cpu_pwr_svi_telemetry_rails (list, optional): value override for args.cpu_pwr_svi_telemetry_rails. Defaults to None
cpu_io_bandwidth (list, optional): value override for args.cpu_io_bandwidth. Defaults to None
cpu_xgmi_bandwidth (list, optional): value override for args.cpu_xgmi_bandwidth. Defaults to None
cpu_metrics_ver (bool, optional): Value override for args.cpu_metrics_ver. Defaults to None
cpu_metrics_table (bool, optional): Value override for args.cpu_metrics_table. Defaults to None
cpu_socket_energy (bool, optional): Value override for args.cpu_socket_energy. Defaults to None
cpu_ddr_bandwidth (bool, optional): Value override for args.cpu_ddr_bandwidth. Defaults to None
cpu_temp (bool, optional): Value override for args.cpu_temp. Defaults to None
cpu_dimm_temp_range_rate (list, optional): Dimm address. Value override for args.cpu_dimm_temp_range_rate. Defaults to None
cpu_dimm_pow_consumption (list, optional): Dimm address. Value override for args.cpu_dimm_pow_consumption. Defaults to None
cpu_dimm_thermal_sensor (list, optional): Dimm address. Value override for args.cpu_dimm_thermal_sensor. Defaults to None
cpu_dfcstate_ctrl (bool, optional): Value override for args.cpu_dfcstate_ctrl. Defaults to None
cpu_railisofreq_policy (bool, optional): Value override for args.cpu_railisofreq_policy. Defaults to None
Returns:
None: Print output via AMDSMILogger to destination
"""
if cpu:
args.cpu = cpu
if cpu_power_metrics:
args.cpu_power_metrics = cpu_power_metrics
if cpu_prochot:
args.cpu_prochot = cpu_prochot
if cpu_freq_metrics:
args.cpu_freq_metrics = cpu_freq_metrics
if cpu_c0_res:
args.cpu_c0_res = cpu_c0_res
if cpu_lclk_dpm_level:
args.cpu_lclk_dpm_level = cpu_lclk_dpm_level
if cpu_pwr_svi_telemetry_rails:
args.cpu_pwr_svi_telemetry_rails = cpu_pwr_svi_telemetry_rails
if cpu_io_bandwidth:
args.cpu_io_bandwidth = cpu_io_bandwidth
if cpu_xgmi_bandwidth:
args.cpu_xgmi_bandwidth = cpu_xgmi_bandwidth
if cpu_metrics_ver:
args.cpu_metrics_ver = cpu_metrics_ver
if cpu_metrics_table:
args.cpu_metrics_table = cpu_metrics_table
if cpu_socket_energy:
args.cpu_socket_energy = cpu_socket_energy
if cpu_ddr_bandwidth:
args.cpu_ddr_bandwidth = cpu_ddr_bandwidth
if cpu_temp:
args.cpu_temp = cpu_temp
if cpu_dimm_temp_range_rate:
args.cpu_dimm_temp_range_rate = cpu_dimm_temp_range_rate
if cpu_dimm_pow_consumption:
args.cpu_dimm_pow_consumption = cpu_dimm_pow_consumption
if cpu_dimm_thermal_sensor:
args.cpu_dimm_thermal_sensor = cpu_dimm_thermal_sensor
if cpu_dfcstate_ctrl:
args.cpu_dfcstate_ctrl = cpu_dfcstate_ctrl
if cpu_railisofreq_policy:
args.cpu_railisofreq_policy = cpu_railisofreq_policy
#store cpu args that are applicable to the current platform
curr_platform_cpu_args = ["cpu_power_metrics", "cpu_prochot", "cpu_freq_metrics",
"cpu_c0_res", "cpu_lclk_dpm_level", "cpu_pwr_svi_telemetry_rails",
"cpu_io_bandwidth", "cpu_xgmi_bandwidth", "cpu_metrics_ver",
"cpu_metrics_table", "cpu_socket_energy", "cpu_ddr_bandwidth",
"cpu_temp", "cpu_dimm_temp_range_rate", "cpu_dimm_pow_consumption",
"cpu_dimm_thermal_sensor", "cpu_dfcstate_ctrl", "cpu_railisofreq_policy"]
curr_platform_cpu_values = [args.cpu_power_metrics, args.cpu_prochot, args.cpu_freq_metrics,
args.cpu_c0_res, args.cpu_lclk_dpm_level, args.cpu_pwr_svi_telemetry_rails,
args.cpu_io_bandwidth, args.cpu_xgmi_bandwidth, args.cpu_metrics_ver,
args.cpu_metrics_table, args.cpu_socket_energy, args.cpu_ddr_bandwidth,
args.cpu_temp, args.cpu_dimm_temp_range_rate, args.cpu_dimm_pow_consumption,
args.cpu_dimm_thermal_sensor, args.cpu_dfcstate_ctrl, args.cpu_railisofreq_policy]
# Handle No CPU passed (fall back as this should be defined in metric())
if args.cpu == None:
args.cpu = self.cpu_handles
if not any(curr_platform_cpu_values):
for arg in curr_platform_cpu_args:
if arg not in("cpu_lclk_dpm_level", "cpu_io_bandwidth", "cpu_xgmi_bandwidth",
"cpu_dimm_temp_range_rate", "cpu_dimm_pow_consumption", "cpu_dimm_thermal_sensor"):
setattr(args, arg, True)
handled_multiple_cpus, device_handle = self.helpers.handle_cpus(args,
self.logger,
self.metric_cpu)
if handled_multiple_cpus:
return # This function is recursive
args.cpu = device_handle
# get cpu id for logging
cpu_id = self.helpers.get_cpu_id_from_device_handle(args.cpu)
logging.debug(f"Metric Arg information for CPU {cpu_id} on {self.helpers.os_info()}")
static_dict = {}
if self.logger.is_json_format():
static_dict['cpu'] = int(cpu_id)
if args.cpu_power_metrics:
static_dict["power_metrics"] = {}
try:
soc_pow = amdsmi_interface.amdsmi_get_cpu_socket_power(args.cpu)
static_dict["power_metrics"]["socket power"] = soc_pow
except amdsmi_exception.AmdSmiLibraryException as e:
static_dict["power_metrics"]["socket power"] = "N/A"
logging.debug("Failed to get socket power for cpu %s | %s", cpu_id, e.get_error_info())
try:
soc_pwr_limit = amdsmi_interface.amdsmi_get_cpu_socket_power_cap(args.cpu)
static_dict["power_metrics"]["socket power limit"] = soc_pwr_limit
except amdsmi_exception.AmdSmiLibraryException as e:
static_dict["power_metrics"]["socket power limit"] = "N/A"
logging.debug("Failed to get socket power limit for cpu %s | %s", cpu_id, e.get_error_info())
try:
soc_max_pwr_limit = amdsmi_interface.amdsmi_get_cpu_socket_power_cap_max(args.cpu)
static_dict["power_metrics"]["socket max power limit"] = soc_max_pwr_limit
except amdsmi_exception.AmdSmiLibraryException as e:
static_dict["power_metrics"]["socket max power limit"] = "N/A"
logging.debug("Failed to get max socket power limit for cpu %s | %s", cpu_id, e.get_error_info())
if args.cpu_prochot:
static_dict["prochot"] = {}
try:
proc_status = amdsmi_interface.amdsmi_get_cpu_prochot_status(args.cpu)
static_dict["prochot"]["prochot_status"] = proc_status
except amdsmi_exception.AmdSmiLibraryException as e:
static_dict["prochot"]["prochot_status"] = "N/A"
logging.debug("Failed to get prochot status for cpu %s | %s", cpu_id, e.get_error_info())
if args.cpu_freq_metrics:
static_dict["freq_metrics"] = {}
try:
fclk_mclk = amdsmi_interface.amdsmi_get_cpu_fclk_mclk(args.cpu)
static_dict["freq_metrics"]["fclkmemclk"] = fclk_mclk
except amdsmi_exception.AmdSmiLibraryException as e:
static_dict["freq_metrics"]["fclkmemclk"] = "N/A"
logging.debug("Failed to get current fclkmemclk freq for cpu %s | %s", cpu_id, e.get_error_info())
try:
cclk_freq = amdsmi_interface.amdsmi_get_cpu_cclk_limit(args.cpu)
static_dict["freq_metrics"]["cclkfreqlimit"] = cclk_freq
except amdsmi_exception.AmdSmiLibraryException as e:
static_dict["freq_metrics"]["cclkfreqlimit"] = "N/A"
logging.debug("Failed to get current cclk freq for cpu %s | %s", cpu_id, e.get_error_info())
try:
soc_cur_freq_limit = amdsmi_interface.amdsmi_get_cpu_socket_current_active_freq_limit(args.cpu)
static_dict["freq_metrics"]["soc_current_active_freq_limit"] = soc_cur_freq_limit
except amdsmi_exception.AmdSmiLibraryException as e:
static_dict["freq_metrics"]["soc_current_active_freq_limit"] = "N/A"
logging.debug("Failed to get socket current freq limit for cpu %s | %s", cpu_id, e.get_error_info())
try:
soc_freq_range = amdsmi_interface.amdsmi_get_cpu_socket_freq_range(args.cpu)
static_dict["freq_metrics"]["soc_freq_range"] = soc_freq_range
except amdsmi_exception.AmdSmiLibraryException as e:
static_dict["freq_metrics"]["soc_freq_range"] = "N/A"
logging.debug("Failed to get socket freq range for cpu %s | %s", cpu_id, e.get_error_info())
if args.cpu_c0_res:
static_dict["c0_residency"] = {}
try:
residency = amdsmi_interface.amdsmi_get_cpu_socket_c0_residency(args.cpu)
static_dict["c0_residency"]["residency"] = residency
except amdsmi_exception.AmdSmiLibraryException as e:
static_dict["c0_residency"]["residency"] = "N/A"
logging.debug("Failed to get C0 residency for cpu %s | %s", cpu_id, e.get_error_info())
if args.cpu_lclk_dpm_level:
static_dict["socket_dpm"] = {}
try:
dpm_val = amdsmi_interface.amdsmi_get_cpu_socket_lclk_dpm_level(args.cpu,
args.cpu_lclk_dpm_level[0][0])
static_dict["socket_dpm"]["dpml_level_range"] = dpm_val
except amdsmi_exception.AmdSmiLibraryException as e:
static_dict["socket_dpm"]["dpml_level_range"] = "N/A"
logging.debug("Failed to get socket dpm level range for cpu %s | %s", cpu_id, e.get_error_info())
if args.cpu_pwr_svi_telemetry_rails:
static_dict["svi_telemetry_all_rails"] = {}
try:
power = amdsmi_interface.amdsmi_get_cpu_pwr_svi_telemetry_all_rails(args.cpu)
static_dict["svi_telemetry_all_rails"]["power"] = power
except amdsmi_exception.AmdSmiLibraryException as e:
static_dict["c0_residency"]["residency"] = "N/A"
logging.debug("Failed to get svi telemetry all rails for cpu %s | %s", cpu_id, e.get_error_info())
if args.cpu_io_bandwidth:
static_dict["io_bandwidth"] = {}
try:
bandwidth = amdsmi_interface.amdsmi_get_cpu_current_io_bandwidth(args.cpu,
int(args.cpu_io_bandwidth[0][0]),
args.cpu_io_bandwidth[0][1].upper())
static_dict["io_bandwidth"]["band_width"] = bandwidth
except amdsmi_exception.AmdSmiLibraryException as e:
static_dict["io_bandwidth"]["band_width"] = "N/A"
logging.debug("Failed to get io bandwidth for cpu %s | %s", cpu_id, e.get_error_info())
if args.cpu_xgmi_bandwidth:
static_dict["xgmi_bandwidth"] = {}
try:
bandwidth = amdsmi_interface.amdsmi_get_cpu_current_xgmi_bw(args.cpu,
int(args.cpu_xgmi_bandwidth[0][0]),
args.cpu_xgmi_bandwidth[0][1].upper())
static_dict["xgmi_bandwidth"]["band_width"] = bandwidth
except amdsmi_exception.AmdSmiLibraryException as e:
static_dict["xgmi_bandwidth"]["band_width"] = "N/A"
logging.debug("Failed to get xgmi bandwidth for cpu %s | %s", cpu_id, e.get_error_info())
if args.cpu_metrics_ver:
static_dict["metric_version"] = {}
try:
version = amdsmi_interface.amdsmi_get_hsmp_metrics_table_version(args.cpu)
static_dict["metric_version"]["version"] = version
except amdsmi_exception.AmdSmiLibraryException as e:
static_dict["metric_version"]["version"] = "N/A"
logging.debug("Failed to get metrics table version for cpu %s | %s", cpu_id, e.get_error_info())
if args.cpu_metrics_table:
static_dict["metrics_table"] = {}
try:
cpu_fam = amdsmi_interface.amdsmi_get_cpu_family()
static_dict["metrics_table"]["cpu_family"] = cpu_fam
except amdsmi_exception.AmdSmiLibraryException as e:
static_dict["metrics_table"]["cpu_family"] = "N/A"
logging.debug("Failed to get cpu family | %s", e.get_error_info())
try:
cpu_mod = amdsmi_interface.amdsmi_get_cpu_model()
static_dict["metrics_table"]["cpu_model"] = cpu_mod
except amdsmi_exception.AmdSmiLibraryException as e:
static_dict["metrics_table"]["cpu_model"] = "N/A"
logging.debug("Failed to get cpu model | %s", e.get_error_info())
try:
cpu_metrics_table = amdsmi_interface.amdsmi_get_hsmp_metrics_table(args.cpu)
static_dict["metrics_table"]["response"] = cpu_metrics_table
except amdsmi_exception.AmdSmiLibraryException as e:
static_dict["metrics_table"]["response"] = "N/A"
logging.debug("Failed to get metrics table for cpu %s | %s", cpu_id, e.get_error_info())
if args.cpu_socket_energy:
static_dict["socket_energy"] = {}
try:
energy = amdsmi_interface.amdsmi_get_cpu_socket_energy(args.cpu)
static_dict["socket_energy"]["response"] = energy
except amdsmi_exception.AmdSmiLibraryException as e:
static_dict["socket_energy"]["response"] = "N/A"
logging.debug("Failed to get socket energy for cpu %s | %s", cpu_id, e.get_error_info())
if args.cpu_ddr_bandwidth:
static_dict["ddr_bandwidth"] = {}
try:
resp = amdsmi_interface.amdsmi_get_cpu_ddr_bw(args.cpu)
static_dict["ddr_bandwidth"]["response"] = resp
except amdsmi_exception.AmdSmiLibraryException as e:
static_dict["ddr_bandwidth"]["response"] = "N/A"
logging.debug("Failed to get ddr bandwdith for cpu %s | %s", cpu_id, e.get_error_info())
if args.cpu_temp:
static_dict["cpu_temp"] = {}
try:
resp = amdsmi_interface.amdsmi_get_cpu_socket_temperature(args.cpu)
static_dict["cpu_temp"]["response"] = resp
except amdsmi_exception.AmdSmiLibraryException as e:
static_dict["cpu_temp"]["response"] = "N/A"
logging.debug("Failed to get cpu temperature for cpu %s | %s", cpu_id, e.get_error_info())
if args.cpu_dimm_temp_range_rate:
static_dict["dimm_temp_range_rate"] = {}
try:
resp = amdsmi_interface.amdsmi_get_cpu_dimm_temp_range_and_refresh_rate(args.cpu, args.cpu_dimm_temp_range_rate[0][0])
static_dict["dimm_temp_range_rate"]["response"] = resp
except amdsmi_exception.AmdSmiLibraryException as e:
static_dict["dimm_temp_range_rate"]["response"] = "N/A"
logging.debug("Failed to get dimm temperature range and refresh rate for cpu %s | %s", cpu_id, e.get_error_info())
if args.cpu_dimm_pow_consumption:
static_dict["dimm_pow_consumption"] = {}
try:
resp = amdsmi_interface.amdsmi_get_cpu_dimm_power_consumption(args.cpu, args.cpu_dimm_pow_consumption[0][0])
static_dict["dimm_pow_consumption"]["response"] = resp
except amdsmi_exception.AmdSmiLibraryException as e:
static_dict["dimm_pow_consumption"]["response"] = "N/A"
logging.debug("Failed to get dimm temperature range and refresh rate for cpu %s | %s", cpu_id, e.get_error_info())
if args.cpu_dimm_thermal_sensor:
static_dict["dimm_thermal_sensor"] = {}
try:
resp = amdsmi_interface.amdsmi_get_cpu_dimm_thermal_sensor(args.cpu, args.cpu_dimm_thermal_sensor[0][0])
static_dict["dimm_thermal_sensor"]["response"] = resp
except amdsmi_exception.AmdSmiLibraryException as e:
static_dict["dimm_thermal_sensor"]["response"] = "N/A"
logging.debug("Failed to get dimm temperature range and refresh rate for cpu %s | %s", cpu_id, e.get_error_info())
if args.cpu_dfcstate_ctrl:
static_dict["dfcstate"] = {}
try:
dfcstatectrl_status = amdsmi_interface.amdsmi_get_dfc_ctrl(args.cpu)
static_dict["dfcstate"]["dfcstatectrl_status"] = dfcstatectrl_status
except amdsmi_exception.AmdSmiLibraryException as e:
static_dict["dfcstate"]["dfcstatectrl_status"] = "N/A"
logging.debug("Failed to get dfcstate control status for cpu %s | %s", cpu_id, e.get_error_info())
if args.cpu_railisofreq_policy:
static_dict["cpurailiso"] = {}
try:
cpurailisofreq_policy = amdsmi_interface.amdsmi_get_cpu_rail_isofreq_policy(args.cpu)
static_dict["cpurailiso"]["cpurailisofreq_policy"] = cpurailisofreq_policy
except amdsmi_exception.AmdSmiLibraryException as e:
static_dict["cpurailiso"]["cpurailisofreq_policy"] = "N/A"
logging.debug("Failed to get cpurailiso frequency policy for cpu %s | %s", cpu_id, e.get_error_info())
multiple_devices_csv_override = False
if not self.logger.is_json_format():
self.logger.store_cpu_output(args.cpu, 'values', static_dict)
else:
self.logger.store_cpu_json_output.append(static_dict)
if multiple_devices:
self.logger.store_multiple_device_output()
return # Skip printing when there are multiple devices
if not self.logger.is_json_format():
self.logger.print_output(multiple_device_enabled=multiple_devices_csv_override)
def metric_core(self, args, multiple_devices=False, core=None, core_boost_limit=None,
core_curr_active_freq_core_limit=None, core_energy=None):
"""Get Static information for target core
Args:
args (Namespace): Namespace containing the parsed CLI args
multiple_devices (bool, optional): True if checking for multiple devices. Defaults to False.
core (device_handle, optional): device_handle for target core. Defaults to None.
core_boost_limit (bool, optional): Value override for args.core_boost_limit. Defaults to None
core_curr_active_freq_core_limit (bool, optional): Value override for args.core_curr_active_freq_core_limit. Defaults to None
core_energy (bool, optional): Value override for args.core_energy. Defaults to None
Returns:
None: Print output via AMDSMILogger to destination
"""
if core:
args.core = core
if core_boost_limit:
args.core_boost_limit = core_boost_limit
if core_curr_active_freq_core_limit:
args.core_curr_active_freq_core_limit = core_curr_active_freq_core_limit
if core_energy:
args.core_energy = core_energy
#store core args that are applicable to the current platform
curr_platform_core_args = ["core_boost_limit", "core_curr_active_freq_core_limit", "core_energy"]
curr_platform_core_values = [args.core_boost_limit, args.core_curr_active_freq_core_limit, args.core_energy]
# Handle No cores passed
if args.core == None:
args.core = self.core_handles
if not any(curr_platform_core_values):
for arg in curr_platform_core_args:
setattr(args, arg, True)
handled_multiple_cores, device_handle = self.helpers.handle_cores(args,
self.logger,
self.metric_core)
if handled_multiple_cores:
return # This function is recursive
args.core = device_handle
# get core id for logging
core_id = self.helpers.get_core_id_from_device_handle(args.core)
logging.debug(f"Static Arg information for Core {core_id} on {self.helpers.os_info()}")
static_dict = {}
if self.logger.is_json_format():
static_dict['core'] = int(core_id)
if args.core_boost_limit:
static_dict["boost_limit"] ={}
try:
core_boost_limit = amdsmi_interface.amdsmi_get_cpu_core_boostlimit(args.core)
static_dict["boost_limit"]["value"] = core_boost_limit
except amdsmi_exception.AmdSmiLibraryException as e:
static_dict["boost_limit"]["value"] = "N/A"
logging.debug("Failed to get core boost limit for core %s | %s", core_id, e.get_error_info())
if args.core_curr_active_freq_core_limit:
static_dict["curr_active_freq_core_limit"] = {}
try:
freq = amdsmi_interface.amdsmi_get_cpu_core_current_freq_limit(args.core)
static_dict["curr_active_freq_core_limit"]["value"] = freq
except amdsmi_exception.AmdSmiLibraryException as e:
static_dict["curr_active_freq_core_limit"]["value"] = "N/A"
logging.debug("Failed to get current active frequency core for core %s | %s", core_id, e.get_error_info())
if args.core_energy:
static_dict["core_energy"] ={}
try:
energy = amdsmi_interface.amdsmi_get_cpu_core_energy(args.core)
static_dict["core_energy"]["value"] = energy
except amdsmi_exception.AmdSmiLibraryException as e:
static_dict["core_energy"]["value"] = "N/A"
logging.debug("Failed to get core energy for core %s | %s", core_id, e.get_error_info())
multiple_devices_csv_override = False
if not self.logger.is_json_format():
self.logger.store_core_output(args.core, 'values', static_dict)
else:
self.logger.store_core_json_output.append(static_dict)
if multiple_devices:
self.logger.store_multiple_device_output()
return # Skip printing when there are multiple devices
if not self.logger.is_json_format():
self.logger.print_output(multiple_device_enabled=multiple_devices_csv_override)
def metric(self, args, multiple_devices=False, watching_output=False, gpu=None,
usage=None, watch=None, watch_time=None, iterations=None, power=None,
clock=None, temperature=None, ecc=None, ecc_blocks=None, pcie=None,
fan=None, voltage_curve=None, overdrive=None, perf_level=None,
xgmi_err=None, energy=None, mem_usage=None, voltage=None, schedule=None,
guard=None, guest_data=None, fb_usage=None, xgmi=None,
cpu=None, cpu_power_metrics=None, cpu_prochot=None, cpu_freq_metrics=None,
cpu_c0_res=None, cpu_lclk_dpm_level=None, cpu_pwr_svi_telemetry_rails=None,
cpu_io_bandwidth=None, cpu_xgmi_bandwidth=None, cpu_metrics_ver=None,
cpu_metrics_table=None, cpu_socket_energy=None, cpu_ddr_bandwidth=None,
cpu_temp=None, cpu_dimm_temp_range_rate=None, cpu_dimm_pow_consumption=None,
cpu_dimm_thermal_sensor=None, cpu_dfcstate_ctrl=None, cpu_railisofreq_policy=None,
core=None, core_boost_limit=None, core_curr_active_freq_core_limit=None,
core_energy=None, throttle=None, base_board=None, gpu_board=None):
"""Get Metric information for target gpu
Args:
args (Namespace): Namespace containing the parsed CLI args
multiple_devices (bool, optional): True if checking for multiple devices. Defaults to False.
watching_output (bool, optional): True if watch argument has been set. Defaults to False.
gpu (device_handle, optional): device_handle for target device. Defaults to None.
usage (bool, optional): Value override for args.usage. Defaults to None.
watch (Positive int, optional): Value override for args.watch. Defaults to None.
watch_time (Positive int, optional): Value override for args.watch_time. Defaults to None.
iterations (Positive int, optional): Value override for args.iterations. Defaults to None.
power (bool, optional): Value override for args.power. Defaults to None.
clock (bool, optional): Value override for args.clock. Defaults to None.
temperature (bool, optional): Value override for args.temperature. Defaults to None.
ecc (bool, optional): Value override for args.ecc. Defaults to None.
ecc_blocks (bool, optional): Value override for args.ecc. Defaults to None.
pcie (bool, optional): Value override for args.pcie. Defaults to None.
fan (bool, optional): Value override for args.fan. Defaults to None.
voltage_curve (bool, optional): Value override for args.voltage_curve. Defaults to None.
overdrive (bool, optional): Value override for args.overdrive. Defaults to None.
perf_level (bool, optional): Value override for args.perf_level. Defaults to None.
xgmi_err (bool, optional): Value override for args.xgmi_err. Defaults to None.
energy (bool, optional): Value override for args.energy. Defaults to None.
mem_usage (bool, optional): Value override for args.mem_usage. Defaults to None.
voltage (bool, optional): Value override for args.voltage. Defaults to None.
schedule (bool, optional): Value override for args.schedule. Defaults to None.
guard (bool, optional): Value override for args.guard. Defaults to None.
guest_data (bool, optional): Value override for args.guest_data. Defaults to None.
fb_usage (bool, optional): Value override for args.fb_usage. Defaults to None.
xgmi (bool, optional): Value override for args.xgmi. Defaults to None.
cpu (cpu_handle, optional): device_handle for target device. Defaults to None.
cpu_power_metrics (bool, optional): Value override for args.cpu_power_metrics. Defaults to None
cpu_prochot (bool, optional): Value override for args.cpu_prochot. Defaults to None.
cpu_freq_metrics (bool, optional): Value override for args.cpu_freq_metrics. Defaults to None.
cpu_c0_res (bool, optional): Value override for args.cpu_c0_res. Defaults to None
cpu_lclk_dpm_level (list, optional): Value override for args.cpu_lclk_dpm_level. Defaults to None
cpu_pwr_svi_telemetry_rails (list, optional): value override for args.cpu_pwr_svi_telemetry_rails. Defaults to None
cpu_io_bandwidth (list, optional): value override for args.cpu_io_bandwidth. Defaults to None
cpu_xgmi_bandwidth (list, optional): value override for args.cpu_xgmi_bandwidth. Defaults to None
cpu_metrics_ver (bool, optional): Value override for args.cpu_metrics_ver. Defaults to None
cpu_metrics_table (bool, optional): Value override for args.cpu_metrics_table. Defaults to None
cpu_socket_energy (bool, optional): Value override for args.cpu_socket_energy. Defaults to None
cpu_ddr_bandwidth (bool, optional): Value override for args.cpu_ddr_bandwidth. Defaults to None
cpu_temp (bool, optional): Value override for args.cpu_temp. Defaults to None
cpu_dimm_temp_range_rate (list, optional): Dimm address. Value override for args.cpu_dimm_temp_range_rate. Defaults to None
cpu_dimm_pow_consumption (list, optional): Dimm address. Value override for args.cpu_dimm_pow_consumption. Defaults to None
cpu_dimm_thermal_sensor (list, optional): Dimm address. Value override for args.cpu_dimm_thermal_sensor. Defaults to None
cpu_dfcstate_ctrl (bool, optional): Value override for args.cpu_dfcstate_ctrl. Defaults to None
cpu_railisofreq_policy (bool, optional): Value override for args.cpu_railisofreq_policy. Defaults to None
core (device_handle, optional): device_handle for target core. Defaults to None.
core_boost_limit (bool, optional): Value override for args.core_boost_limit. Defaults to None
core_curr_active_freq_core_limit (bool, optional): Value override for args.core_curr_active_freq_core_limit. Defaults to None
core_energy (bool, optional): Value override for args.core_energy. Defaults to None
Raises:
IndexError: Index error if gpu list is empty
Returns:
None: Print output via AMDSMILogger to destination
"""
# TODO Move watch logic into here and make it driver agnostic or enable it for CPU arguments
# Mutually exclusive args
if gpu:
args.gpu = gpu
if cpu:
args.cpu = cpu
if core:
args.core = core
# Check if a GPU argument has been set
gpu_args_enabled = False
gpu_attributes = ["usage", "watch", "watch_time", "iterations", "power", "clock",
"temperature", "ecc", "ecc_blocks", "pcie", "fan", "voltage_curve",
"overdrive", "perf_level", "xgmi_err", "energy", "mem_usage", "voltage", "schedule",
"guard", "guest_data", "fb_usage", "xgmi", "throttle", "base_board", "gpu_board"]
for attr in gpu_attributes:
if hasattr(args, attr):
if getattr(args, attr):
gpu_args_enabled = True
break
# Check if a CPU argument has been set
cpu_args_enabled = False
cpu_attributes = ["cpu_power_metrics", "cpu_prochot", "cpu_freq_metrics", "cpu_c0_res",
"cpu_lclk_dpm_level", "cpu_pwr_svi_telemetry_rails", "cpu_io_bandwidth",
"cpu_xgmi_bandwidth", "cpu_metrics_ver", "cpu_metrics_table",
"cpu_socket_energy", "cpu_ddr_bandwidth", "cpu_temp", "cpu_dimm_temp_range_rate",
"cpu_dimm_pow_consumption", "cpu_dimm_thermal_sensor",
"cpu_dfcstate_ctrl", "cpu_railisofreq_policy"]
for attr in cpu_attributes:
if hasattr(args, attr):
if getattr(args, attr):
cpu_args_enabled = True
break
# Check if a Core argument has been set
core_args_enabled = False
core_attributes = ["core_boost_limit", "core_curr_active_freq_core_limit", "core_energy"]
for attr in core_attributes:
if hasattr(args, attr):
if getattr(args, attr):
core_args_enabled = True
break
# Handle CPU and GPU driver intialization cases
if self.helpers.is_amd_hsmp_initialized() and self.helpers.is_amdgpu_initialized():
logging.debug("gpu_args_enabled: %s, cpu_args_enabled: %s, core_args_enabled: %s",
gpu_args_enabled, cpu_args_enabled, core_args_enabled)
logging.debug("args.gpu: %s, args.cpu: %s, args.core: %s", args.gpu, args.cpu, args.core)
# If a GPU or CPU argument is provided only print out the specified device.
if args.cpu == None and args.gpu == None and args.core == None:
# If no args are set, print out all CPU, GPU, and Core metrics info
if not gpu_args_enabled and not cpu_args_enabled and not core_args_enabled:
args.cpu = self.cpu_handles
args.gpu = self.device_handles
args.core = self.core_handles
# Handle cases where the user has only specified an argument and no specific device
if args.gpu == None and gpu_args_enabled:
args.gpu = self.device_handles
if args.cpu == None and cpu_args_enabled:
args.cpu = self.cpu_handles
if args.core == None and core_args_enabled:
args.core = self.core_handles
# Print out CPU first
if args.cpu:
self.metric_cpu(args, multiple_devices, cpu, cpu_power_metrics, cpu_prochot,
cpu_freq_metrics, cpu_c0_res, cpu_lclk_dpm_level,
cpu_pwr_svi_telemetry_rails, cpu_io_bandwidth, cpu_xgmi_bandwidth,
cpu_metrics_ver, cpu_metrics_table, cpu_socket_energy,
cpu_ddr_bandwidth, cpu_temp, cpu_dimm_temp_range_rate,
cpu_dimm_pow_consumption, cpu_dimm_thermal_sensor,
cpu_dfcstate_ctrl, cpu_railisofreq_policy)
if args.core:
self.logger.output = {}
self.logger.clear_multiple_devices_output()
self.metric_core(args, multiple_devices, core, core_boost_limit,
core_curr_active_freq_core_limit, core_energy)
if args.gpu:
self.logger.output = {}
self.logger.clear_multiple_devices_output()
self.metric_gpu(args, multiple_devices, watching_output, gpu,
usage, watch, watch_time, iterations, power,
clock, temperature, ecc, ecc_blocks, pcie,
fan, voltage_curve, overdrive, perf_level,
xgmi_err, energy, mem_usage, voltage, schedule,
guard, guest_data, fb_usage, xgmi, throttle,
base_board, gpu_board)
elif self.helpers.is_amd_hsmp_initialized(): # Only CPU is initialized
if args.cpu == None and args.core == None:
# If no args are set, print out all CPU and Core metrics info
if not cpu_args_enabled and not core_args_enabled:
args.cpu = self.cpu_handles
args.core = self.core_handles
if args.cpu == None and cpu_args_enabled:
args.cpu = self.cpu_handles
if args.core == None and core_args_enabled:
args.core = self.core_handles
if args.cpu:
self.metric_cpu(args, multiple_devices, cpu, cpu_power_metrics, cpu_prochot,
cpu_freq_metrics, cpu_c0_res, cpu_lclk_dpm_level,
cpu_pwr_svi_telemetry_rails, cpu_io_bandwidth, cpu_xgmi_bandwidth,
cpu_metrics_ver, cpu_metrics_table, cpu_socket_energy,
cpu_ddr_bandwidth, cpu_temp, cpu_dimm_temp_range_rate,
cpu_dimm_pow_consumption, cpu_dimm_thermal_sensor,
cpu_dfcstate_ctrl, cpu_railisofreq_policy)
if args.core:
self.logger.output = {}
self.logger.clear_multiple_devices_output()
self.metric_core(args, multiple_devices, core, core_boost_limit,
core_curr_active_freq_core_limit, core_energy)
elif self.helpers.is_amdgpu_initialized(): # Only GPU is initialized
if args.gpu == None:
args.gpu = self.device_handles
self.logger.clear_multiple_devices_output()
self.metric_gpu(args, multiple_devices, watching_output, gpu,
usage, watch, watch_time, iterations, power,
clock, temperature, ecc, ecc_blocks, pcie,
fan, voltage_curve, overdrive, perf_level,
xgmi_err, energy, mem_usage, voltage, schedule, throttle,
base_board, gpu_board)
if self.logger.is_json_format():
self.logger.combine_arrays_to_json()
def process(self, args, multiple_devices=False, watching_output=False,
gpu=None, general=None, engine=None, pid=None, name=None,
watch=None, watch_time=None, iterations=None):
"""Get Process Information from the target GPU
Args:
args (Namespace): Namespace containing the parsed CLI args
multiple_devices (bool, optional): True if checking for multiple devices. Defaults to False.
watching_output (bool, optional): True if watch argument has been set. Defaults to False.
gpu (device_handle, optional): device_handle for target device. Defaults to None.
general (bool, optional): Value override for args.general. Defaults to None.
engine (bool, optional): Value override for args.engine. Defaults to None.
pid (Positive int, optional): Value override for args.pid. Defaults to None.
name (str, optional): Value override for args.name. Defaults to None.
watch (Positive int, optional): Value override for args.watch. Defaults to None.
watch_time (Positive int, optional): Value override for args.watch_time. Defaults to None.
iterations (Positive int, optional): Value override for args.iterations. Defaults to None.
Raises:
IndexError: Index error if gpu list is empty
Returns:
None: Print output via AMDSMILogger to destination
"""
# Set args.* to passed in arguments
if gpu:
args.gpu = gpu
if general:
args.general = general
if engine:
args.engine = engine
if pid:
args.pid = pid
if name:
args.name = name
if watch:
args.watch = watch
if watch_time:
args.watch_time = watch_time
if iterations:
args.iterations = iterations
# Handle No GPU passed
if args.gpu == None:
args.gpu = self.device_handles
# Handle watch logic, will only enter this block once
if args.watch:
self.helpers.handle_watch(args=args, subcommand=self.process, logger=self.logger)
return
# Handle multiple GPUs
if isinstance(args.gpu, list):
if len(args.gpu) > 1:
# Deepcopy gpus as recursion will destroy the gpu list
stored_gpus = []
for gpu in args.gpu:
stored_gpus.append(gpu)
# Store output from multiple devices
for device_handle in args.gpu:
self.process(args, multiple_devices=True, watching_output=watching_output, gpu=device_handle)
# Reload original gpus
args.gpu = stored_gpus
# Print multiple device output
self.logger.print_output(multiple_device_enabled=True, watching_output=watching_output)
# Add output to total watch output and clear multiple device output
if watching_output:
self.logger.store_watch_output(multiple_device_enabled=True)
# Flush the watching output
self.logger.print_output(multiple_device_enabled=True, watching_output=watching_output)
return
elif len(args.gpu) == 1:
args.gpu = args.gpu[0]
else:
raise IndexError("args.gpu should not be an empty list")
# Get gpu_id for logging
gpu_id = self.helpers.get_gpu_id_from_device_handle(args.gpu)
# Populate initial processes
try:
process_list = amdsmi_interface.amdsmi_get_gpu_process_list(args.gpu)
except amdsmi_exception.AmdSmiLibraryException as e:
logging.debug("Failed to get process list for gpu %s | %s", gpu_id, e.get_error_info())
raise e
filtered_process_values = []
for process_info in process_list:
process_info = {
"name": process_info["name"],
"pid": process_info["pid"],
"memory_usage": {
"gtt_mem": process_info["memory_usage"]["gtt_mem"],
"cpu_mem": process_info["memory_usage"]["cpu_mem"],
"vram_mem": process_info["memory_usage"]["vram_mem"],
},
"mem_usage": process_info["mem"],
"usage": {
"gfx": process_info["engine_usage"]["gfx"],
"enc": process_info["engine_usage"]["enc"],
},
"cu_occupancy": process_info["cu_occupancy"],
"evicted_time": process_info["evicted_time"]
}
engine_usage_unit = "ns"
memory_usage_unit = "B"
evicted_time_unit = "ms"
if self.logger.is_human_readable_format():
process_info['mem_usage'] = self.helpers.convert_bytes_to_readable(process_info['mem_usage'])
for usage_metric in process_info['memory_usage']:
process_info["memory_usage"][usage_metric] = self.helpers.convert_bytes_to_readable(process_info["memory_usage"][usage_metric])
memory_usage_unit = ""
process_info['mem_usage'] = self.helpers.unit_format(self.logger,
process_info['mem_usage'],
memory_usage_unit)
process_info['evicted_time'] = self.helpers.unit_format(self.logger,
process_info['evicted_time'],
evicted_time_unit)
for usage_metric in process_info['usage']:
process_info['usage'][usage_metric] = self.helpers.unit_format(self.logger,
process_info['usage'][usage_metric],
engine_usage_unit)
for usage_metric in process_info['memory_usage']:
process_info['memory_usage'][usage_metric] = self.helpers.unit_format(self.logger,
process_info['memory_usage'][usage_metric],
memory_usage_unit)
filtered_process_values.append({'process_info': process_info})
if not filtered_process_values:
process_info = "N/A"
logging.debug("Failed to detect any process on gpu %s", gpu_id)
filtered_process_values.append({'process_info': process_info})
# Arguments will filter the populated processes
# General and Engine to expose process_info values
if args.general or args.engine:
for process_info in filtered_process_values:
if not process_info['process_info'] == "N/A":
if args.general and args.engine:
del process_info['process_info']['memory_usage']
elif args.general:
del process_info['process_info']['memory_usage']
del process_info['process_info']['usage'] # Used in engine
elif args.engine:
del process_info['process_info']['memory_usage']
del process_info['process_info']['mem_usage'] # Used in general
# Filter out non specified pids
if args.pid:
process_pids = []
for process_info in filtered_process_values:
if process_info['process_info'] == "N/A":
continue
pid = str(process_info['process_info']['pid'])
if str(args.pid) == pid:
process_pids.append(process_info)
filtered_process_values = process_pids
# Filter out non specified process names
if args.name:
process_names = []
for process_info in filtered_process_values:
if process_info['process_info'] == "N/A":
continue
process_name = str(process_info['process_info']['name']).lower()
if str(args.name).lower() == process_name:
process_names.append(process_info)
filtered_process_values = process_names
# If the name or pid args filter processes out then insert an N/A placeholder
if not filtered_process_values:
filtered_process_values.append({'process_info': "N/A"})
logging.debug(f"Process Info for GPU {gpu_id} | {filtered_process_values}")
for index, process in enumerate(filtered_process_values):
if process['process_info'] == "N/A":
filtered_process_values[index]['process_info'] = "No running processes detected"
if self.logger.is_json_format():
if watching_output:
self.logger.store_output(args.gpu, 'timestamp', int(time.time()))
self.logger.store_output(args.gpu, 'process_list', filtered_process_values)
if self.logger.is_human_readable_format():
if watching_output:
self.logger.store_output(args.gpu, 'timestamp', int(time.time()))
# When we print out process_info we remove the index
# The removal is needed only for human readable process format to align with Host
for index, process in enumerate(filtered_process_values):
self.logger.store_output(args.gpu, f'process_info_{index}', process['process_info'])
multiple_devices_csv_override = False
if self.logger.is_csv_format():
multiple_devices_csv_override = True
for process in filtered_process_values:
if watching_output:
self.logger.store_output(args.gpu, 'timestamp', int(time.time()))
self.logger.store_output(args.gpu, 'process_info', process['process_info'])
self.logger.store_multiple_device_output()
if multiple_devices:
self.logger.store_multiple_device_output()
return # Skip printing when there are multiple devices
multiple_devices = multiple_devices or multiple_devices_csv_override
self.logger.print_output(multiple_device_enabled=multiple_devices, watching_output=watching_output)
if watching_output: # End of single gpu add to watch_output
self.logger.store_watch_output(multiple_device_enabled=multiple_devices)
def profile(self, args):
"""Not applicable to linux baremetal"""
print('Not applicable to linux baremetal')
def event(self, args, gpu=None):
""" Get event information for target gpus
Args:
args (Namespace): argparser args to pass to subcommand
gpu (device_handle, optional): device_handle for target device. Defaults to None.
Return:
stdout event information for target gpus
"""
if args.gpu:
gpu = args.gpu
if gpu == None:
args.gpu = self.device_handles
if not isinstance(args.gpu, list):
args.gpu = [args.gpu]
print('EVENT LISTENING:\n')
print('Press q and hit ENTER when you want to stop.')
self.stop = False
threads = []
for device_handle in range(len(args.gpu)):
x = threading.Thread(target=self._event_thread, args=(self, device_handle))
threads.append(x)
x.start()
previous_sigterm_handler = signal.getsignal(signal.SIGTERM)
system_exit_exc = None
signal.signal(signal.SIGTERM, self._event_sigterm_handler)
try:
while True:
try:
user_input = input()
except EOFError:
self.stop = True
break
except KeyboardInterrupt:
self.stop = True
break
if self.stop:
break
if user_input == 'q':
print("Escape Sequence Detected; Exiting")
self.stop = True
break
except SystemExit as exc:
system_exit_exc = exc
finally:
self.stop = True
for thread in threads:
thread.join()
signal.signal(signal.SIGTERM, previous_sigterm_handler)
if system_exit_exc is not None:
raise system_exit_exc
def _event_sigterm_handler(self, signum, frame):
self.stop = True
raise SystemExit(128 + signum)
def topology(self, args, multiple_devices=False, gpu=None, access=None,
weight=None, hops=None, link_type=None, numa_bw=None,
coherent=None, atomics=None, dma=None, bi_dir=None):
""" Get topology information for target gpus
params:
args - argparser args to pass to subcommand
multiple_devices (bool) - True if checking for multiple devices
gpu (device_handle) - device_handle for target device
access (bool) - Value override for args.access
weight (bool) - Value override for args.weight
hops (bool) - Value override for args.hops
type (bool) - Value override for args.type
numa_bw (bool) - Value override for args.numa_bw
coherent (bool) - Value override for args.coherent
atomics (bool) - Value override for args.atomics
dma (bool) - Value override for args.dma
bi_dir (bool) - Value override for args.bi_dir
return:
Nothing
"""
# Set args.* to passed in arguments
if gpu:
args.gpu = gpu
if access:
args.access = access
if weight:
args.weight = weight
if hops:
args.hops = hops
if link_type:
args.link_type = link_type
if numa_bw:
args.numa_bw = numa_bw
if coherent:
args.coherent = coherent
if atomics:
args.atomics = atomics
if dma:
args.dma = dma
if bi_dir:
args.bi_dir = bi_dir
# Handle No GPU passed
if args.gpu == None:
args.gpu = self.device_handles
if not isinstance(args.gpu, list):
args.gpu = [args.gpu]
# Handle all args being false
if not any([args.access, args.weight, args.hops, args.link_type, args.numa_bw,
args.coherent, args.atomics, args.dma, args.bi_dir]):
args.access = args.weight = args.hops = args.link_type= args.numa_bw = \
args.coherent = args.atomics = args.dma = args.bi_dir = True
# Clear the table header
self.logger.table_header = ''.rjust(12)
if not self.group_check_printed:
self.helpers.check_required_groups(check_render=True, check_video=False)
self.group_check_printed = True
p2p_status_cache = {}
def get_cached_p2p_status(src_gpu, dest_gpu):
#Get P2P status with caching to avoid duplicate calls
src_gpu_id = self.helpers.get_gpu_id_from_device_handle(src_gpu)
dest_gpu_id = self.helpers.get_gpu_id_from_device_handle(dest_gpu)
key = (src_gpu_id, dest_gpu_id)
if key not in p2p_status_cache:
try:
if src_gpu == dest_gpu:
p2p_status_cache[key] = {"cap": {
"is_iolink_coherent": -1,
"is_iolink_atomics_32bit": -1,
"is_iolink_atomics_64bit": -1,
"is_iolink_dma": -1,
"is_iolink_bi_directional": -1
}}
else:
p2p_status_cache[key] = amdsmi_interface.amdsmi_topo_get_p2p_status(src_gpu, dest_gpu)
except amdsmi_exception.AmdSmiLibraryException as e:
logging.debug("Failed to get link status for %s to %s | %s",
src_gpu_id,
dest_gpu_id,
e.get_error_info())
p2p_status_cache[key] ={
"cap":
{
"is_iolink_coherent": -1,
"is_iolink_atomics_32bit": -1,
"is_iolink_atomics_64bit": -1,
"is_iolink_dma": -1,
"is_iolink_bi_directional": -1
}
}
return p2p_status_cache[key]
# Populate the possible gpus
topo_values = []
for src_gpu_index, src_gpu in enumerate(args.gpu):
src_gpu_id = self.helpers.get_gpu_id_from_device_handle(src_gpu)
topo_values.append({"gpu" : src_gpu_id})
src_gpu_bdf = amdsmi_interface.amdsmi_get_gpu_device_bdf(src_gpu)
topo_values[src_gpu_index]['bdf'] = src_gpu_bdf
self.logger.table_header += src_gpu_bdf.rjust(13)
if not self.logger.is_json_format():
continue # below is for JSON format only
##########################
# JSON formatting start #
##########################
links = []
# create json obj for data alignment
# dest_gpu_links = {
# "gpu": GPU #
# "bdf": BDF identification
# "weight": 0 - self (current node); weight >= 0 correlated with hops (GPU-CPU, GPU-GPU, GPU-CPU-CPU-GPU, etc..)
# "link_status": "ENABLED" - devices linked; "DISABLED" - devices not linked; Correlated to access
# "link_type": "SELF" - current node, "PCIE", "XGMI", "N/A" - no link,"UNKNOWN" - unidentified link type
# "num_hops": num_hops - # of hops between devices
# "bandwidth": numa_bw - The NUMA "minimum bandwidth-maximum bandwidth" beween src and dest nodes
# "N/A" - self node or not connected devices
# "coherent": coherent - Coherant / Non-Coherant io links
# "atomics": atomics - 32 and 64-bit atomic io link capability between nodes
# "dma": dma - P2P direct memory access (DMA) link capability between nodes
# "bi_dir": bi_dir - P2P bi-directional link capability between nodes
# }
for dest_gpu_index, dest_gpu in enumerate(args.gpu):
link_type = "SELF"
if src_gpu != dest_gpu:
link_type = amdsmi_interface.amdsmi_topo_get_link_type(src_gpu, dest_gpu)['type']
if isinstance(link_type, int):
if link_type == amdsmi_interface.amdsmi_wrapper.AMDSMI_LINK_TYPE_INTERNAL:
link_type = "UNKNOWN"
elif link_type == amdsmi_interface.amdsmi_wrapper.AMDSMI_LINK_TYPE_PCIE:
link_type = "PCIE"
elif link_type == amdsmi_interface.amdsmi_wrapper.AMDSMI_LINK_TYPE_XGMI:
link_type = "XGMI"
else:
link_type = "N/A"
numa_bw = "N/A"
if src_gpu != dest_gpu:
try:
bw_dict = amdsmi_interface.amdsmi_get_minmax_bandwidth_between_processors(src_gpu, dest_gpu)
numa_bw = f"{bw_dict['min_bandwidth']}-{bw_dict['max_bandwidth']}"
except amdsmi_exception.AmdSmiLibraryException as e:
logging.debug("Failed to get min max bandwidth for %s to %s | %s",
self.helpers.get_gpu_id_from_device_handle(src_gpu),
self.helpers.get_gpu_id_from_device_handle(dest_gpu),
e.get_error_info())
weight = 0
num_hops = 0
if src_gpu != dest_gpu:
weight = amdsmi_interface.amdsmi_topo_get_link_weight(src_gpu, dest_gpu)
num_hops = amdsmi_interface.amdsmi_topo_get_link_type(src_gpu, dest_gpu)['hops']
link_status = amdsmi_interface.amdsmi_is_P2P_accessible(src_gpu, dest_gpu)
if link_status:
link_status = "ENABLED"
else:
link_status = "DISABLED"
link_coherent = "SELF"
link_atomics = "SELF"
link_dma = "SELF"
link_bi_dir = "SELF"
if src_gpu != dest_gpu:
try:
cap = get_cached_p2p_status(src_gpu, dest_gpu)['cap']
link_coherent = (
"C" if cap['is_iolink_coherent'] == 1 else
"NC" if cap['is_iolink_coherent'] == 0 else
"N/A"
)
link_atomics = (
"64,32" if cap['is_iolink_atomics_32bit'] == 1 and cap['is_iolink_atomics_64bit'] == 1 else
"32" if cap['is_iolink_atomics_32bit'] == 1 else
"64" if cap['is_iolink_atomics_64bit'] == 1 else
"N/A"
)
link_dma = (
"T" if cap['is_iolink_dma'] == 1 else
"F" if cap['is_iolink_dma'] == 0 else
"N/A"
)
link_bi_dir = (
"T" if cap['is_iolink_bi_directional'] == 1 else
"F" if cap['is_iolink_bi_directional'] == 0 else
"N/A"
)
except amdsmi_exception.AmdSmiLibraryException as e:
logging.debug("Failed to get link status for %s to %s | %s",
self.helpers.get_gpu_id_from_device_handle(src_gpu),
self.helpers.get_gpu_id_from_device_handle(dest_gpu),
e.get_error_info())
# link_status = amdsmi_is_P2P_accessible(src,dest)
dest_gpu_links = {
"gpu": self.helpers.get_gpu_id_from_device_handle(dest_gpu),
"bdf": amdsmi_interface.amdsmi_get_gpu_device_bdf(dest_gpu),
"weight": weight,
"link_status": link_status,
"link_type": link_type,
"num_hops": num_hops,
"bandwidth": numa_bw,
"coherent": link_coherent,
"atomics": link_atomics,
"dma": link_dma,
"bi_dir": link_bi_dir
}
if not args.access:
del dest_gpu_links['link_status']
if not args.weight:
del dest_gpu_links['weight']
if not args.link_type:
del dest_gpu_links['link_type']
if not args.hops:
del dest_gpu_links['num_hops']
if not args.numa_bw:
del dest_gpu_links['bandwidth']
if not args.coherent:
del dest_gpu_links['coherent']
if not args.atomics:
del dest_gpu_links['atomics']
if not args.dma:
del dest_gpu_links['dma']
if not args.bi_dir:
del dest_gpu_links['bi_dir']
links.append(dest_gpu_links)
dest_end = dest_gpu_index+1 == len(args.gpu)
isEndOfSrc = src_gpu_index+1 == len(args.gpu)
if dest_end:
topo_values[src_gpu_index]['links'] = links
continue
if isEndOfSrc:
self.logger.multiple_device_output = topo_values
self.logger.print_output(multiple_device_enabled=True, tabular=True)
return
##########################
# JSON formatting end #
##########################
if args.access:
tabular_output = []
for src_gpu_index, src_gpu in enumerate(args.gpu):
src_gpu_bdf = amdsmi_interface.amdsmi_get_gpu_device_bdf(src_gpu)
if self.logger.is_human_readable_format():
tabular_output_dict = {'gpu' : f"{src_gpu_bdf} "}
else:
tabular_output_dict = {'gpu' : src_gpu_bdf}
src_gpu_links = {}
for dest_gpu in args.gpu:
dest_gpu_id = self.helpers.get_gpu_id_from_device_handle(dest_gpu)
dest_gpu_key = f'gpu_{dest_gpu_id}'
try:
dest_gpu_link_status = amdsmi_interface.amdsmi_is_P2P_accessible(src_gpu, dest_gpu)
if dest_gpu_link_status:
src_gpu_links[dest_gpu_key] = "ENABLED"
else:
src_gpu_links[dest_gpu_key] = "DISABLED"
except amdsmi_exception.AmdSmiLibraryException as e:
src_gpu_links[dest_gpu_key] = "N/A"
logging.debug("Failed to get link status for %s to %s | %s",
self.helpers.get_gpu_id_from_device_handle(src_gpu),
self.helpers.get_gpu_id_from_device_handle(dest_gpu),
e.get_error_info())
topo_values[src_gpu_index]['link_accessibility'] = src_gpu_links
tabular_output_dict.update(src_gpu_links)
tabular_output.append(tabular_output_dict)
if self.logger.is_human_readable_format():
self.logger.multiple_device_output = tabular_output
self.logger.table_title = "ACCESS TABLE"
self.logger.print_output(multiple_device_enabled=True, tabular=True)
if args.weight:
tabular_output = []
for src_gpu_index, src_gpu in enumerate(args.gpu):
src_gpu_bdf = amdsmi_interface.amdsmi_get_gpu_device_bdf(src_gpu)
if self.logger.is_human_readable_format():
tabular_output_dict = {'gpu' : f"{src_gpu_bdf} "}
else:
tabular_output_dict = {'gpu' : src_gpu_bdf}
src_gpu_weight = {}
for dest_gpu in args.gpu:
dest_gpu_id = self.helpers.get_gpu_id_from_device_handle(dest_gpu)
dest_gpu_key = f'gpu_{dest_gpu_id}'
if src_gpu == dest_gpu:
src_gpu_weight[dest_gpu_key] = 0
continue
try:
dest_gpu_link_weight = amdsmi_interface.amdsmi_topo_get_link_weight(src_gpu, dest_gpu)
src_gpu_weight[dest_gpu_key] = dest_gpu_link_weight
except amdsmi_exception.AmdSmiLibraryException as e:
src_gpu_weight[dest_gpu_key] = "N/A"
logging.debug("Failed to get link weight for %s to %s | %s",
self.helpers.get_gpu_id_from_device_handle(src_gpu),
self.helpers.get_gpu_id_from_device_handle(dest_gpu),
e.get_error_info())
topo_values[src_gpu_index]['weight'] = src_gpu_weight
tabular_output_dict.update(src_gpu_weight)
tabular_output.append(tabular_output_dict)
if self.logger.is_human_readable_format():
self.logger.multiple_device_output = tabular_output
self.logger.table_title = "WEIGHT TABLE"
self.logger.print_output(multiple_device_enabled=True, tabular=True)
if args.hops:
tabular_output = []
for src_gpu_index, src_gpu in enumerate(args.gpu):
src_gpu_bdf = amdsmi_interface.amdsmi_get_gpu_device_bdf(src_gpu)
if self.logger.is_human_readable_format():
tabular_output_dict = {'gpu' : f"{src_gpu_bdf} "}
else:
tabular_output_dict = {'gpu' : src_gpu_bdf}
src_gpu_hops = {}
for dest_gpu in args.gpu:
dest_gpu_id = self.helpers.get_gpu_id_from_device_handle(dest_gpu)
dest_gpu_key = f'gpu_{dest_gpu_id}'
if src_gpu == dest_gpu:
src_gpu_hops[dest_gpu_key] = 0
continue
try:
dest_gpu_hops = amdsmi_interface.amdsmi_topo_get_link_type(src_gpu, dest_gpu)['hops']
src_gpu_hops[dest_gpu_key] = dest_gpu_hops
except amdsmi_exception.AmdSmiLibraryException as e:
src_gpu_hops[dest_gpu_key] = "N/A"
logging.debug("Failed to get link hops for %s to %s | %s",
self.helpers.get_gpu_id_from_device_handle(src_gpu),
self.helpers.get_gpu_id_from_device_handle(dest_gpu),
e.get_error_info())
topo_values[src_gpu_index]['hops'] = src_gpu_hops
tabular_output_dict.update(src_gpu_hops)
tabular_output.append(tabular_output_dict)
if self.logger.is_human_readable_format():
self.logger.multiple_device_output = tabular_output
self.logger.table_title = "HOPS TABLE"
self.logger.print_output(multiple_device_enabled=True, tabular=True)
if args.link_type:
tabular_output = []
for src_gpu_index, src_gpu in enumerate(args.gpu):
src_gpu_bdf = amdsmi_interface.amdsmi_get_gpu_device_bdf(src_gpu)
if self.logger.is_human_readable_format():
tabular_output_dict = {'gpu' : f"{src_gpu_bdf} "}
else:
tabular_output_dict = {'gpu' : src_gpu_bdf}
src_gpu_link_type = {}
for dest_gpu in args.gpu:
dest_gpu_id = self.helpers.get_gpu_id_from_device_handle(dest_gpu)
dest_gpu_key = f'gpu_{dest_gpu_id}'
if src_gpu == dest_gpu:
src_gpu_link_type[dest_gpu_key] = "SELF"
continue
try:
link_type = amdsmi_interface.amdsmi_topo_get_link_type(src_gpu, dest_gpu)['type']
if isinstance(link_type, int):
if link_type == amdsmi_interface.amdsmi_wrapper.AMDSMI_LINK_TYPE_INTERNAL:
src_gpu_link_type[dest_gpu_key] = "UNKNOWN"
elif link_type == amdsmi_interface.amdsmi_wrapper.AMDSMI_LINK_TYPE_PCIE:
src_gpu_link_type[dest_gpu_key] = "PCIE"
elif link_type == amdsmi_interface.amdsmi_wrapper.AMDSMI_LINK_TYPE_XGMI:
src_gpu_link_type[dest_gpu_key] = "XGMI"
except amdsmi_exception.AmdSmiLibraryException as e:
src_gpu_link_type[dest_gpu_key] = "N/A"
logging.debug("Failed to get link type for %s to %s | %s",
self.helpers.get_gpu_id_from_device_handle(src_gpu),
self.helpers.get_gpu_id_from_device_handle(dest_gpu),
e.get_error_info())
topo_values[src_gpu_index]['link_type'] = src_gpu_link_type
tabular_output_dict.update(src_gpu_link_type)
tabular_output.append(tabular_output_dict)
if self.logger.is_human_readable_format():
self.logger.multiple_device_output = tabular_output
self.logger.table_title = "LINK TYPE TABLE"
self.logger.print_output(multiple_device_enabled=True, tabular=True)
if args.numa_bw:
tabular_output = []
for src_gpu_index, src_gpu in enumerate(args.gpu):
src_gpu_bdf = amdsmi_interface.amdsmi_get_gpu_device_bdf(src_gpu)
if self.logger.is_human_readable_format():
tabular_output_dict = {'gpu' : f"{src_gpu_bdf} "}
else:
tabular_output_dict = {'gpu' : src_gpu_bdf}
src_gpu_link_type = {}
for dest_gpu in args.gpu:
dest_gpu_id = self.helpers.get_gpu_id_from_device_handle(dest_gpu)
dest_gpu_key = f'gpu_{dest_gpu_id}'
if src_gpu == dest_gpu:
src_gpu_link_type[dest_gpu_key] = "N/A"
continue
try:
link_type = amdsmi_interface.amdsmi_topo_get_link_type(src_gpu, dest_gpu)['type']
if isinstance(link_type, int):
if link_type != amdsmi_interface.amdsmi_wrapper.AMDSMI_LINK_TYPE_XGMI:
# non_xgmi = True
src_gpu_link_type[dest_gpu_key] = "N/A"
continue
except amdsmi_exception.AmdSmiLibraryException as e:
src_gpu_link_type[dest_gpu_key] = "N/A"
logging.debug("Failed to get link type for %s to %s | %s",
self.helpers.get_gpu_id_from_device_handle(src_gpu),
self.helpers.get_gpu_id_from_device_handle(dest_gpu),
e.get_error_info())
try:
bw_dict = amdsmi_interface.amdsmi_get_minmax_bandwidth_between_processors(src_gpu, dest_gpu)
src_gpu_link_type[dest_gpu_key] = f"{bw_dict['min_bandwidth']}-{bw_dict['max_bandwidth']}"
except amdsmi_exception.AmdSmiLibraryException as e:
src_gpu_link_type[dest_gpu_key] = e.get_error_info()
logging.debug("Failed to get min max bandwidth for %s to %s | %s",
self.helpers.get_gpu_id_from_device_handle(src_gpu),
self.helpers.get_gpu_id_from_device_handle(dest_gpu),
e.get_error_info())
topo_values[src_gpu_index]['numa_bandwidth'] = src_gpu_link_type
tabular_output_dict.update(src_gpu_link_type)
tabular_output.append(tabular_output_dict)
if self.logger.is_human_readable_format():
self.logger.multiple_device_output = tabular_output
self.logger.table_title = "NUMA BW TABLE"
self.logger.print_output(multiple_device_enabled=True, tabular=True)
if args.coherent:
tabular_output = []
for src_gpu_index, src_gpu in enumerate(args.gpu):
src_gpu_bdf = amdsmi_interface.amdsmi_get_gpu_device_bdf(src_gpu)
if self.logger.is_human_readable_format():
tabular_output_dict = {'gpu' : f"{src_gpu_bdf} "}
else:
tabular_output_dict = {'gpu' : src_gpu_bdf}
src_gpu_coherent = {}
for dest_gpu in args.gpu:
dest_gpu_id = self.helpers.get_gpu_id_from_device_handle(dest_gpu)
dest_gpu_key = f'gpu_{dest_gpu_id}'
if src_gpu == dest_gpu:
src_gpu_coherent[dest_gpu_key] = "SELF"
continue
try:
iolink_coherent = get_cached_p2p_status(src_gpu, dest_gpu)['cap']['is_iolink_coherent']
src_gpu_coherent[dest_gpu_key] = "C" if iolink_coherent == 1 else "NC" if iolink_coherent == 0 else "N/A"
except amdsmi_exception.AmdSmiLibraryException as e:
src_gpu_coherent[dest_gpu_key] = "N/A"
logging.debug("Failed to get link coherent for %s to %s | %s",
self.helpers.get_gpu_id_from_device_handle(src_gpu),
self.helpers.get_gpu_id_from_device_handle(dest_gpu),
e.get_error_info())
topo_values[src_gpu_index]['coherent'] = src_gpu_coherent
tabular_output_dict.update(src_gpu_coherent)
tabular_output.append(tabular_output_dict)
if self.logger.is_human_readable_format():
self.logger.multiple_device_output = tabular_output
self.logger.table_title = "CACHE COHERANCY TABLE"
self.logger.print_output(multiple_device_enabled=True, tabular=True)
if args.atomics:
tabular_output = []
for src_gpu_index, src_gpu in enumerate(args.gpu):
src_gpu_bdf = amdsmi_interface.amdsmi_get_gpu_device_bdf(src_gpu)
if self.logger.is_human_readable_format():
tabular_output_dict = {'gpu' : f"{src_gpu_bdf} "}
else:
tabular_output_dict = {'gpu' : src_gpu_bdf}
src_gpu_atomics = {}
for dest_gpu in args.gpu:
dest_gpu_id = self.helpers.get_gpu_id_from_device_handle(dest_gpu)
dest_gpu_key = f'gpu_{dest_gpu_id}'
if src_gpu == dest_gpu:
src_gpu_atomics[dest_gpu_key] = "SELF"
continue
try:
cap = get_cached_p2p_status(src_gpu, dest_gpu)['cap']
src_gpu_atomics[dest_gpu_key] = (
"64,32" if cap['is_iolink_atomics_32bit'] == 1 and cap['is_iolink_atomics_64bit'] == 1 else
"32" if cap['is_iolink_atomics_32bit'] == 1 else
"64" if cap['is_iolink_atomics_64bit'] == 1 else
"N/A"
)
except amdsmi_exception.AmdSmiLibraryException as e:
src_gpu_atomics[dest_gpu_key] = "N/A"
logging.debug("Failed to get link atomics for %s to %s | %s",
self.helpers.get_gpu_id_from_device_handle(src_gpu),
self.helpers.get_gpu_id_from_device_handle(dest_gpu),
e.get_error_info())
topo_values[src_gpu_index]['atomics'] = src_gpu_atomics
tabular_output_dict.update(src_gpu_atomics)
tabular_output.append(tabular_output_dict)
if self.logger.is_human_readable_format():
self.logger.multiple_device_output = tabular_output
self.logger.table_title = "ATOMICS TABLE"
self.logger.print_output(multiple_device_enabled=True, tabular=True)
if args.dma:
tabular_output = []
for src_gpu_index, src_gpu in enumerate(args.gpu):
src_gpu_bdf = amdsmi_interface.amdsmi_get_gpu_device_bdf(src_gpu)
if self.logger.is_human_readable_format():
tabular_output_dict = {'gpu' : f"{src_gpu_bdf} "}
else:
tabular_output_dict = {'gpu' : src_gpu_bdf}
src_gpu_dma = {}
for dest_gpu in args.gpu:
dest_gpu_id = self.helpers.get_gpu_id_from_device_handle(dest_gpu)
dest_gpu_key = f'gpu_{dest_gpu_id}'
if src_gpu == dest_gpu:
src_gpu_dma[dest_gpu_key] = "SELF"
continue
try:
iolink_dma = get_cached_p2p_status(src_gpu, dest_gpu)['cap']['is_iolink_dma']
src_gpu_dma[dest_gpu_key] = "T" if iolink_dma == 1 else "F" if iolink_dma == 0 else "N/A"
except amdsmi_exception.AmdSmiLibraryException as e:
src_gpu_dma[dest_gpu_key] = "N/A"
logging.debug("Failed to get link dma for %s to %s | %s",
self.helpers.get_gpu_id_from_device_handle(src_gpu),
self.helpers.get_gpu_id_from_device_handle(dest_gpu),
e.get_error_info())
topo_values[src_gpu_index]['dma'] = src_gpu_dma
tabular_output_dict.update(src_gpu_dma)
tabular_output.append(tabular_output_dict)
if self.logger.is_human_readable_format():
self.logger.multiple_device_output = tabular_output
self.logger.table_title = "DMA TABLE"
self.logger.print_output(multiple_device_enabled=True, tabular=True)
if args.bi_dir:
tabular_output = []
for src_gpu_index, src_gpu in enumerate(args.gpu):
src_gpu_bdf = amdsmi_interface.amdsmi_get_gpu_device_bdf(src_gpu)
if self.logger.is_human_readable_format():
tabular_output_dict = {'gpu' : f"{src_gpu_bdf} "}
else:
tabular_output_dict = {'gpu' : src_gpu_bdf}
src_gpu_bi_dir = {}
for dest_gpu in args.gpu:
dest_gpu_id = self.helpers.get_gpu_id_from_device_handle(dest_gpu)
dest_gpu_key = f'gpu_{dest_gpu_id}'
if src_gpu == dest_gpu:
src_gpu_bi_dir[dest_gpu_key] = "SELF"
continue
try:
iolink_bi_dir = get_cached_p2p_status(src_gpu, dest_gpu)['cap']['is_iolink_bi_directional']
src_gpu_bi_dir[dest_gpu_key] = "T" if iolink_bi_dir == 1 else "F" if iolink_bi_dir == 0 else "N/A"
except amdsmi_exception.AmdSmiLibraryException as e:
src_gpu_bi_dir[dest_gpu_key] = "N/A"
logging.debug("Failed to get link bi-directional for %s to %s | %s",
self.helpers.get_gpu_id_from_device_handle(src_gpu),
self.helpers.get_gpu_id_from_device_handle(dest_gpu),
e.get_error_info())
topo_values[src_gpu_index]['bi_dir'] = src_gpu_bi_dir
tabular_output_dict.update(src_gpu_bi_dir)
tabular_output.append(tabular_output_dict)
if self.logger.is_human_readable_format():
self.logger.multiple_device_output = tabular_output
self.logger.table_title = "BI-DIRECTIONAL TABLE"
self.logger.print_output(multiple_device_enabled=True, tabular=True)
if self.logger.is_human_readable_format():
# Populate the legend output
legend_parts = [
"\n\nLegend:",
" SELF = Current GPU",
" ENABLED / DISABLED = Link is enabled or disabled",
" N/A = Not supported",
" T/F = True / False",
" C/NC = Coherant / Non-Coherant io links",
" 64,32 = 64 bit and 32 bit atomic support",
" <BW from>-<BW to>"
]
legend_output = "\n".join(legend_parts)
if self.logger.destination == 'stdout':
print(legend_output)
else:
with self.logger.destination.open('a', encoding="utf-8") as output_file:
output_file.write(legend_output + '\n')
self.logger.multiple_device_output = topo_values
if self.logger.is_csv_format():
new_output = []
for elem in self.logger.multiple_device_output:
new_output.append(self.logger.flatten_dict(elem, topology_override=True))
self.logger.multiple_device_output = new_output
if not self.logger.is_human_readable_format():
self.logger.print_output(multiple_device_enabled=True)
def set_core(self, args, multiple_devices=False, core=None, core_boost_limit=None):
"""Issue set commands to target core(s)
Args:
args (Namespace): Namespace containing the parsed CLI args
multiple_devices (bool, optional): True if checking for multiple devices. Defaults to False.
core (device_handle, optional): device_handle for target device. Defaults to None.
core_boost_limit (list, optional): Value override for args.core_boost_limit. Defaults to None. Defaults to None.
Raises:
ValueError: Value error if no core value is provided
IndexError: Index error if core list is empty
Return:
Nothing
"""
if core:
args.core = core
if core_boost_limit:
args.core_boost_limit = core_boost_limit
if args.core == None:
raise ValueError('No Core provided, specific Core targets(S) are needed')
# Handle multiple cores
handled_multiple_cores, device_handle = self.helpers.handle_cores(args, self.logger, self.set_core)
if handled_multiple_cores:
return # This function is recursive
# Error if no subcommand args are passed
if not any([args.core_boost_limit]):
command = " ".join(sys.argv[1:])
raise AmdSmiRequiredCommandException(command, self.logger.format)
args.core = device_handle
# build core string for errors
try:
core_id = self.helpers.get_core_id_from_device_handle(args.core)
except IndexError:
core_id = f'ID Unavailable for {args.core}'
static_dict = {}
if args.core_boost_limit:
static_dict["set_core_boost_limit"] = {}
try:
amdsmi_interface.amdsmi_set_cpu_core_boostlimit(args.core, args.core_boost_limit[0][0])
#Verify the core boost limit is set
boost_limit = amdsmi_interface.amdsmi_get_cpu_core_boostlimit(args.core)
# Extract numeric value from response (remove units if present)
if isinstance(boost_limit, str):
# Extract just the number part (assumes format like "5000 MHz" or "5000")
boost_limit = int(boost_limit.split()[0])
else:
boost_limit = int(boost_limit)
if boost_limit < args.core_boost_limit[0][0]:
static_dict["set_core_boost_limit"]["Response"] = f"Max allowed boostlimit is {boost_limit} MHz"
elif boost_limit > args.core_boost_limit[0][0]:
static_dict["set_core_boost_limit"]["Response"] = f"Min allowed boostlimit is {boost_limit} MHz"
else:
static_dict["set_core_boost_limit"]["Response"] = f"{boost_limit} MHz"
except amdsmi_exception.AmdSmiLibraryException as e:
static_dict["set_core_boost_limit"]["Response"] = f"Error occurred for Core {core_id} - {e.get_error_info()}"
logging.debug("Failed to set core boost limit for core %s | %s", core_id, e.get_error_info())
multiple_devices_csv_override = False
self.logger.store_core_output(args.core, 'values', static_dict)
if multiple_devices:
self.logger.store_multiple_device_output()
return # Skip printing when there are multiple devices
self.logger.print_output(multiple_device_enabled=multiple_devices_csv_override)
def set_cpu(self, args, multiple_devices=False, cpu=None, cpu_pwr_limit=None,
cpu_xgmi_link_width=None, cpu_lclk_dpm_level=None, cpu_pwr_eff_mode=None,
cpu_gmi3_link_width=None, cpu_pcie_link_rate=None, cpu_df_pstate_range=None,
cpu_enable_apb=None, cpu_disable_apb=None, soc_boost_limit=None,
cpu_dfcstate_ctrl=None, cpu_railisofreq_policy=None):
"""Issue set commands to target cpu(s)
Args:
args (Namespace): Namespace containing the parsed CLI args
multiple_devices (bool, optional): True if checking for multiple devices. Defaults to False.
cpu (cpu_handle, optional): device_handle for target device. Defaults to None.
cpu_pwr_limit (int, optional): Value override for args.cpu_pwr_limit. Defaults to None.
cpu_xgmi_link_width (List[int], optional): Value override for args.cpu_xgmi_link_width. Defaults to None.
cpu_lclk_dpm_level (List[int], optional): Value override for args.cpu_lclk_dpm_level. Defaults to None.
cpu_pwr_eff_mode (int, optional): Value override for args.cpu_pwr_eff_mode. Defaults to None.
cpu_gmi3_link_width (List[int], optional): Value override for args.cpu_gmi3_link_width. Defaults to None.
cpu_pcie_link_rate (int, optional): Value override for args.cpu_pcie_link_rate. Defaults to None.
cpu_df_pstate_range (List[int], optional): Value override for args.cpu_df_pstate_range. Defaults to None.
cpu_enable_apb (bool, optional): Value override for args.cpu_enable_apb. Defaults to None.
cpu_disable_apb (int, optional): Value override for args.cpu_disable_apb. Defaults to None.
soc_boost_limit (int, optional): Value override for args.soc_boost_limit. Defaults to None.
cpu_dfcstate_ctrl (int, optional): Value override for args.cpu_dfcstate_ctrl. Defaults to None.
cpu_railisofreq_policy (int, optional): Value override for args.cpu_railisofreq_policy. Defaults to None.
Raises:
ValueError: Value error if no cpu value is provided
IndexError: Index error if cpu list is empty
Return:
Nothing
"""
if cpu:
args.cpu = cpu
if cpu_pwr_limit:
args.cpu_pwr_limit = cpu_pwr_limit
if cpu_xgmi_link_width:
args.cpu_xgmi_link_width = cpu_xgmi_link_width
if cpu_lclk_dpm_level:
args.cpu_lclk_dpm_level = cpu_lclk_dpm_level
if cpu_pwr_eff_mode:
args.cpu_pwr_eff_mode = cpu_pwr_eff_mode
if cpu_gmi3_link_width:
args.cpu_gmi3_link_width = cpu_gmi3_link_width
if cpu_pcie_link_rate:
args.cpu_pcie_link_rate = cpu_pcie_link_rate
if cpu_df_pstate_range:
args.cpu_df_pstate_range = cpu_df_pstate_range
if cpu_enable_apb:
args.cpu_enable_apb = cpu_enable_apb
if cpu_disable_apb:
args.cpu_disable_apb = cpu_disable_apb
if soc_boost_limit:
args.soc_boost_limit = soc_boost_limit
if cpu_dfcstate_ctrl:
args.cpu_dfcstate_ctrl = cpu_dfcstate_ctrl
if cpu_railisofreq_policy:
args.cpu_railisofreq_policy = cpu_railisofreq_policy
if args.cpu == None:
raise ValueError('No CPU provided, specific CPU targets(S) are needed')
#Handle multiple CPU's
handled_multiple_cpus, device_handle = self.helpers.handle_cpus(args, self.logger, self.set_cpu)
if handled_multiple_cpus:
return # This function is recursive
args.cpu = device_handle
#Error if no subcommand args are passed
if not any([args.cpu_pwr_limit, args.cpu_xgmi_link_width, args.cpu_lclk_dpm_level,
args.cpu_pwr_eff_mode, args.cpu_gmi3_link_width, args.cpu_pcie_link_rate,
args.cpu_df_pstate_range, args.cpu_enable_apb, args.cpu_disable_apb,
args.soc_boost_limit, args.cpu_dfcstate_ctrl, args.cpu_railisofreq_policy]):
command = " ".join(sys.argv[1:])
raise AmdSmiRequiredCommandException(command, self.logger.format)
# Build CPU string for errors
try:
cpu_id = self.helpers.get_cpu_id_from_device_handle(args.cpu)
except IndexError:
cpu_id = f'ID Unavailable for {args.cpu}'
static_dict = {}
if args.cpu_pwr_limit:
static_dict["set_pwr_limit"] = {}
try:
soc_max_pwr_limit = amdsmi_interface.amdsmi_get_cpu_socket_power_cap_max(args.cpu)
extract_numeric = soc_max_pwr_limit.split()[0]
max_power = int(extract_numeric)
amdsmi_interface.amdsmi_set_cpu_socket_power_cap(args.cpu, args.cpu_pwr_limit[0][0])
if args.cpu_pwr_limit[0][0] > max_power:
args.cpu_pwr_limit[0][0] = max_power
static_dict["set_pwr_limit"]["Response"] = f"{args.cpu_pwr_limit[0][0] / 1000:.3f} mW"
except amdsmi_exception.AmdSmiLibraryException as e:
static_dict["set_pwr_limit"]["Response"] = f"Error occurred for CPU {cpu_id} - {e.get_error_info()}"
logging.debug("Failed to set power limit for cpu %s | %s", cpu_id, e.get_error_info())
if args.cpu_xgmi_link_width:
static_dict["set_xgmi_link_width"] = {}
try:
amdsmi_interface.amdsmi_set_cpu_xgmi_width(args.cpu, args.cpu_xgmi_link_width[0][0],
args.cpu_xgmi_link_width[0][1])
static_dict["set_xgmi_link_width"]["Response"] = f"{args.cpu_xgmi_link_width[0][0]} - {args.cpu_xgmi_link_width[0][1]}"
except amdsmi_exception.AmdSmiLibraryException as e:
static_dict["set_xgmi_link_width"]["Response"] = f"Error occurred for CPU {cpu_id} - {e.get_error_info()}"
logging.debug("Failed to set xgmi link width for cpu %s | %s", cpu_id, e.get_error_info())
if args.cpu_lclk_dpm_level:
static_dict["set_lclk_dpm_level"] = {}
try:
amdsmi_interface.amdsmi_set_cpu_socket_lclk_dpm_level(args.cpu, args.cpu_lclk_dpm_level[0][0],
args.cpu_lclk_dpm_level[0][1],
args.cpu_lclk_dpm_level[0][2])
static_dict["set_lclk_dpm_level"]["Response"] = f"NBIO[{args.cpu_lclk_dpm_level[0][0]}]"
except amdsmi_exception.AmdSmiLibraryException as e:
static_dict["set_lclk_dpm_level"]["Response"] = f"Error occurred for CPU {cpu_id} - {e.get_error_info()}"
logging.debug("Failed to set lclk dpm level for cpu %s | %s", cpu_id, e.get_error_info())
if args.cpu_pwr_eff_mode:
static_dict["set_pwr_eff_mode"] = {}
try:
amdsmi_interface.amdsmi_set_cpu_pwr_efficiency_mode(args.cpu, args.cpu_pwr_eff_mode[0][0])
static_dict["set_pwr_eff_mode"]["Response"] = f"{args.cpu_pwr_eff_mode[0][0]}"
except amdsmi_exception.AmdSmiLibraryException as e:
static_dict["set_pwr_eff_mode"]["Response"] = f"Error occurred for CPU {cpu_id} - {e.get_error_info()}"
logging.debug("Failed to set power efficiency mode for cpu %s | %s", cpu_id, e.get_error_info())
if args.cpu_gmi3_link_width:
static_dict["set_gmi3_link_width"] = {}
try:
amdsmi_interface.amdsmi_set_cpu_gmi3_link_width_range(args.cpu, args.cpu_gmi3_link_width[0][0],
args.cpu_gmi3_link_width[0][1])
static_dict["set_gmi3_link_width"]["response"] = f"{args.cpu_gmi3_link_width[0][0]} - {args.cpu_gmi3_link_width[0][1]}"
except amdsmi_exception.AmdSmiLibraryException as e:
static_dict["set_gmi3_link_width"]["response"] = f"Error occurred for CPU {cpu_id} - {e.get_error_info()}"
logging.debug("Failed to set gmi3 link width for cpu %s | %s", cpu_id, e.get_error_info())
if args.cpu_pcie_link_rate:
static_dict["set_pcie_link_rate"] = {}
try:
resp = amdsmi_interface.amdsmi_set_cpu_pcie_link_rate(args.cpu, args.cpu_pcie_link_rate[0][0])
static_dict["set_pcie_link_rate"]["prev_mode"] = resp
except amdsmi_exception.AmdSmiLibraryException as e:
static_dict["set_pcie_link_rate"]["prev_mode"] = f"Error occurred for CPU {cpu_id} - {e.get_error_info()}"
logging.debug("Failed to set pcie link rate for cpu %s | %s", cpu_id, e.get_error_info())
if args.cpu_df_pstate_range:
static_dict["set_df_pstate_range"] = {}
try:
amdsmi_interface.amdsmi_set_cpu_df_pstate_range(args.cpu, args.cpu_df_pstate_range[0][0],
args.cpu_df_pstate_range[0][1])
static_dict["set_df_pstate_range"]["response"] = "Set Operation successful"
except amdsmi_exception.AmdSmiLibraryException as e:
static_dict["set_df_pstate_range"]["response"] = f"Error occurred for CPU {cpu_id} - {e.get_error_info()}"
logging.debug("Failed to set df pstate range for cpu %s | %s", cpu_id, e.get_error_info())
if args.cpu_enable_apb:
static_dict["apbenable"] = {}
try:
amdsmi_interface.amdsmi_cpu_apb_enable(args.cpu)
static_dict["apbenable"]["state"] = "Enabled DF - Pstate performance boost algorithm"
except amdsmi_exception.AmdSmiLibraryException as e:
static_dict["apbenable"]["state"] = f"Error occurred for CPU {cpu_id} - {e.get_error_info()}"
logging.debug("Failed to enable APB for cpu %s | %s", cpu_id, e.get_error_info())
if args.cpu_disable_apb:
static_dict["apbdisable"] = {}
try:
amdsmi_interface.amdsmi_cpu_apb_disable(args.cpu, args.cpu_disable_apb[0][0])
static_dict["apbdisable"]["state"] = "Disabled DF - Pstate performance boost algorithm"
except amdsmi_exception.AmdSmiLibraryException as e:
static_dict["apbdisable"]["state"] = f"Error occurred for CPU {cpu_id} - {e.get_error_info()}"
logging.debug("Failed to enable APB for cpu %s | %s", cpu_id, e.get_error_info())
if args.soc_boost_limit:
static_dict["set_soc_boost_limit"] = {}
try:
amdsmi_interface.amdsmi_set_cpu_socket_boostlimit(args.cpu, args.soc_boost_limit[0][0])
static_dict["set_soc_boost_limit"]["Response"] = "Set Operation successful"
except amdsmi_exception.AmdSmiLibraryException as e:
#static_dict["set_soc_boost_limit"]["Response"] = "N/A"
static_dict["set_soc_boost_limit"]["Response"] = f"Error occurred for CPU {cpu_id} - {e.get_error_info()}"
logging.debug("Failed to set socket boost limit for cpu %s | %s", cpu_id, e.get_error_info())
if args.cpu_dfcstate_ctrl:
static_dict["dfcstatectrl"] = {}
try:
amdsmi_interface.amdsmi_set_dfc_ctrl(args.cpu, args.cpu_dfcstate_ctrl[0][0])
static_dict["dfcstatectrl"]["state"] = "DFCState control operation successful"
except amdsmi_exception.AmdSmiLibraryException as e:
static_dict["dfcstatectrl"]["state"] = f"Error occurred for CPU {cpu_id} - {e.get_error_info()}"
logging.debug("Failed to set dfcstate control for cpu %s | %s", cpu_id, e.get_error_info())
if args.cpu_railisofreq_policy:
static_dict["cpurailiso"] = {}
try:
amdsmi_interface.amdsmi_set_cpu_rail_isofreq_policy(args.cpu, args.cpu_railisofreq_policy[0][0])
static_dict["cpurailiso"]["state"] = "Set CPU ISO frequency policy operation successful"
except amdsmi_exception.AmdSmiLibraryException as e:
static_dict["cpurailiso"]["state"] = f"Error occurred for CPU {cpu_id} - {e.get_error_info()}"
logging.debug("Failed to set ISO frequency policy for cpu %s | %s", cpu_id, e.get_error_info())
multiple_devices_csv_override = False
self.logger.store_cpu_output(args.cpu, 'values', static_dict)
if multiple_devices:
self.logger.store_multiple_device_output()
return # Skip printing when there are multiple devices
self.logger.print_output(multiple_device_enabled=multiple_devices_csv_override)
def set_gpu(self, args, multiple_devices=False, gpu=None, fan=None, perf_level=None,
profile=None, perf_determinism=None, compute_partition=None,
memory_partition=None, power_cap=None, soc_pstate=None, xgmi_plpd = None,
process_isolation=None, clk_limit=None, clk_level=None, ptl_status=None, ptl_format=None):
"""Issue reset commands to target gpu(s)
Args:
args (Namespace): Namespace containing the parsed CLI args
multiple_devices (bool, optional): True if checking for multiple devices. Defaults to False.
gpu (device_handle, optional): device_handle for target device. Defaults to None.
fan (int, optional): Value override for args.fan. Defaults to None.
perf_level (amdsmi_interface.AmdSmiDevPerfLevel, optional): Value override for args.perf_level. Defaults to None.
profile (bool, optional): Value override for args.profile. Defaults to None.
perf_determinism (int, optional): Value override for args.perf_determinism. Defaults to None.
compute_partition (amdsmi_interface.AmdSmiComputePartitionType, optional): Value override for args.compute_partition. Defaults to None.
memory_partition (amdsmi_interface.AmdSmiMemoryPartitionType, optional): Value override for args.memory_partition. Defaults to None.
power_cap (int, optional): Value override for args.power_cap. Defaults to None.
soc_pstate (int, optional): Value override for args.soc_pstate. Defaults to None.
xgmi_plpd (int, optional): Value override for args.xgmi_plpd. Defaults to None.
process_isolation (int, optional): Value override for args.process_isolation. Defaults to None.
ptl_status (int, optional): Value override for args.ptl_status. Defaults to None.
ptl_format(string, optional): Value override for args.ptl_format. Defaults to None.
Raises:
ValueError: Value error if no gpu value is provided
IndexError: Index error if gpu list is empty
Return:
Nothing
"""
# Set args.* to passed in arguments
if gpu:
args.gpu = gpu
if fan is not None:
args.fan = fan
if perf_level:
args.perf_level = perf_level
if profile:
args.profile = profile
if perf_determinism is not None:
args.perf_determinism = perf_determinism
if compute_partition:
args.compute_partition = compute_partition
if memory_partition:
args.memory_partition = memory_partition
if power_cap:
args.power_cap = power_cap
if soc_pstate:
args.soc_pstate = soc_pstate
if xgmi_plpd:
args.xgmi_plpd = xgmi_plpd
if process_isolation:
args.process_isolation = process_isolation
if clk_limit:
args.clk_limit = clk_limit
if clk_level:
args.clk_level = clk_level
if ptl_status:
args.ptl_status = ptl_status
if ptl_format:
args.ptl_format = ptl_format
# Handle No GPU passed
if args.gpu == None:
args.gpu = self.device_handles
if not self.group_check_printed:
self.helpers.check_required_groups(check_render=True, check_video=False)
self.group_check_printed = True
# Handle multiple GPUs
handled_multiple_gpus, device_handle = self.helpers.handle_gpus(args, self.logger, self.set_gpu)
if handled_multiple_gpus:
return # This function is recursive
args.gpu = device_handle
# Error if no subcommand args are passed
if self.helpers.is_baremetal():
if not any([getattr(args, 'fan', None) is not None,
getattr(args, 'perf_level', None) is not None,
getattr(args, 'profile', None) is not None,
getattr(args, 'compute_partition', None) is not None,
getattr(args, 'memory_partition', None) is not None,
getattr(args, 'perf_determinism', None) is not None,
getattr(args, 'power_cap', None) is not None,
getattr(args, 'soc_pstate', None) is not None,
getattr(args, 'xgmi_plpd', None) is not None,
getattr(args, 'clk_level', None) is not None,
getattr(args, 'clk_limit', None) is not None,
getattr(args, 'ptl_status', None) is not None,
getattr(args, 'ptl_format', None) is not None,
getattr(args, 'process_isolation', None) is not None]):
command = " ".join(sys.argv[1:])
raise AmdSmiRequiredCommandException(command, self.logger.format)
else:
if not any([getattr(args, 'power_cap', None) is not None,
getattr(args, 'clk_limit', None) is not None,
getattr(args, 'process_isolation', None) is not None]):
command = " ".join(sys.argv[1:])
raise AmdSmiRequiredCommandException(command, self.logger.format)
# Build GPU string for errors
try:
gpu_bdf = amdsmi_interface.amdsmi_get_gpu_device_bdf(args.gpu)
except amdsmi_exception.AmdSmiLibraryException:
gpu_bdf = f'BDF Unavailable for {args.gpu}'
try:
gpu_id = self.helpers.get_gpu_id_from_device_handle(args.gpu)
except IndexError:
gpu_id = f'ID Unavailable for {args.gpu}'
gpu_string = f"GPU ID: {gpu_id} BDF:{gpu_bdf}"
# Handle args
if self.helpers.is_baremetal():
if isinstance(args.fan, int):
# Convert fan speed to percentage
# Note: amdsmi_set_gpu_fan_speed expects fan speed in RPM, so
# we convert the value to a percentage based on the maximum fan speed of 255 RPM.
# We need to round down the user's passed fan speed % to the nearest whole number.
# This allows us to match the float -> int conversion when converting from percentage to RPM (as previously passed by the parser).
fan_percentage = int((int(args.fan) / 255) * 100 // 1) # round down (aka floor) to nearest whole number
try:
amdsmi_interface.amdsmi_set_gpu_fan_speed(args.gpu, 0, args.fan)
except amdsmi_exception.AmdSmiLibraryException as e:
if e.get_error_code() == amdsmi_interface.amdsmi_wrapper.AMDSMI_STATUS_NO_PERM:
raise PermissionError('Command requires elevation') from e
result = f"[{e.get_error_info(detailed=False)}] Unable to set fan speed to {args.fan} RPM ({fan_percentage}%)"
self.logger.store_output(args.gpu, 'fan', result)
self.logger.print_output()
self.logger.clear_multiple_devices_output()
return
self.logger.store_output(args.gpu, 'fan', f"Successfully set fan speed to {args.fan} RPM ({fan_percentage}%)")
self.logger.print_output()
self.logger.clear_multiple_devices_output()
return
if args.perf_level:
perf_level = amdsmi_interface.AmdSmiDevPerfLevel[args.perf_level]
try:
amdsmi_interface.amdsmi_set_gpu_perf_level(args.gpu, perf_level)
except amdsmi_exception.AmdSmiLibraryException as e:
if e.get_error_code() == amdsmi_interface.amdsmi_wrapper.AMDSMI_STATUS_NO_PERM:
raise PermissionError('Command requires elevation') from e
self.logger.store_output(args.gpu, 'perflevel', f"[{e.get_error_info(detailed=False)}] Unable to set performance level to {args.perf_level}")
perf_options = str(self.helpers.get_perf_levels()[0][0:-1]).replace("[", "").replace("]", "").replace("'", "").replace(" ", "")
print(f"\nPerformance Level Options:\n\t{perf_options}\n")
self.logger.print_output()
self.logger.clear_multiple_devices_output()
return
self.logger.store_output(args.gpu, 'perflevel', f"Successfully set performance level {args.perf_level}")
self.logger.print_output()
self.logger.clear_multiple_devices_output()
return
if args.profile:
try:
# Parse profile input (name or number)
profile_input = args.profile.upper()
name_mapping = self.helpers.get_power_profile_name_mapping()
if profile_input in name_mapping:
profile_mask = name_mapping[profile_input]
else:
# Invalid profile - show available ones
try:
profile_status = amdsmi_interface.amdsmi_get_gpu_power_profile_presets(args.gpu, 0)
available = self.helpers.parse_available_profiles(profile_status['available_profiles'])
available_str = ", ".join(available)
except amdsmi_exception.AmdSmiLibraryException as e:
available_str = "Unable to fetch available profiles"
logging.debug(f"Failed to fetch available profiles: {e.get_error_info()}")
self.logger.store_output(args.gpu, 'profile',
f"Invalid profile: {args.profile}\n\nAvailable profiles: {available_str}")
self.logger.print_output()
self.logger.clear_multiple_devices_output()
return
# Set the profile
amdsmi_interface.amdsmi_set_gpu_power_profile(args.gpu, 0, profile_mask)
self.logger.store_output(args.gpu, 'profile',
f"Successfully set power profile to {profile_input}")
except amdsmi_exception.AmdSmiLibraryException as e:
if e.get_error_code() == amdsmi_interface.amdsmi_wrapper.AMDSMI_STATUS_NO_PERM:
raise PermissionError('Command requires elevation') from e
# Get available profiles for error message
try:
profile_status = amdsmi_interface.amdsmi_get_gpu_power_profile_presets(args.gpu, 0)
available = self.helpers.parse_available_profiles(profile_status['available_profiles'])
available_str = ", ".join(available)
except amdsmi_exception.AmdSmiLibraryException as get_error:
available_str = "Unable to fetch available profiles"
logging.debug(f"Failed to fetch available profiles: {get_error.get_error_info()}")
error_msg = f"[{e.get_error_info(detailed=False)}] Unable to set power profile to {args.profile}"
self.logger.store_output(args.gpu, 'profile', error_msg)
print(f"\nAvailable Power Profiles:\n\t{available_str}\n")
self.logger.print_output()
self.logger.clear_multiple_devices_output()
return
self.logger.print_output()
self.logger.clear_multiple_devices_output()
return
if isinstance(args.perf_determinism, int):
try:
amdsmi_interface.amdsmi_set_gpu_perf_determinism_mode(args.gpu, args.perf_determinism)
except amdsmi_exception.AmdSmiLibraryException as e:
if e.get_error_code() == amdsmi_interface.amdsmi_wrapper.AMDSMI_STATUS_NO_PERM:
raise PermissionError('Command requires elevation') from e
self.logger.store_output(args.gpu, 'perfdeterminism', f"[{e.get_error_info(detailed=False)}] Unable to enable performance determinism and set GFX clock frequency to {args.perf_determinism} MHz")
self.logger.print_output()
self.logger.clear_multiple_devices_output()
return
self.logger.store_output(args.gpu, 'perfdeterminism', f"Successfully enabled performance determinism and set GFX clock frequency to {args.perf_determinism} MHz")
self.logger.print_output()
self.logger.clear_multiple_devices_output()
return
if args.compute_partition:
current_set_count = self.helpers.get_set_count()
future_set_count = 0
attempted_to_set = "N/A"
user_requested_partition_args = "N/A"
try:
(accelerator_set_choices, accelerator_profiles) = self.helpers.get_accelerator_choices_types_indices()
logging.debug("args.compute_partition: %s; Accelerator_set_choices: %s", str(args.compute_partition), str(json.dumps(accelerator_set_choices, indent=4)))
if args.compute_partition in accelerator_profiles['profile_types']:
compute_partition = amdsmi_interface.AmdSmiComputePartitionType[args.compute_partition]
index = accelerator_profiles['profile_types'].index(args.compute_partition)
attempted_to_set = f"Attempted to set accelerator partition to {args.compute_partition} (profile #{accelerator_profiles['profile_indices'][int(index)]}) on {gpu_string}"
user_requested_partition_args = f"{args.compute_partition} (profile #{accelerator_profiles['profile_indices'][int(index)]})"
amdsmi_interface.amdsmi_set_gpu_compute_partition(args.gpu, compute_partition)
elif args.compute_partition in accelerator_profiles['profile_indices']:
compute_partition = int(args.compute_partition)
index = accelerator_profiles['profile_indices'].index(args.compute_partition)
attempted_to_set = f"Attempted to set accelerator partition to {accelerator_profiles['profile_types'][int(index)]} (profile #{args.compute_partition}) on {gpu_string}"
user_requested_partition_args = f"{accelerator_profiles['profile_types'][int(index)]} (profile #{args.compute_partition})"
amdsmi_interface.amdsmi_set_gpu_accelerator_partition_profile(args.gpu, compute_partition)
else:
raise ValueError(f"Invalid accelerator configuration {args.compute_partition} on {gpu_string}")
self.helpers.increment_set_count()
future_set_count = self.helpers.get_set_count()
if current_set_count == future_set_count-1:
self.logger.store_output(args.gpu, 'accelerator_partition', f"Successfully set accelerator partition to {user_requested_partition_args}")
self.logger.print_output()
self.logger.clear_multiple_devices_output()
return
except amdsmi_exception.AmdSmiLibraryException as e:
if e.get_error_code() == amdsmi_interface.amdsmi_wrapper.AMDSMI_STATUS_NO_PERM:
raise PermissionError('Command requires elevation') from e
elif e.get_error_code() == amdsmi_interface.amdsmi_wrapper.AMDSMI_STATUS_NOT_SUPPORTED:
self.helpers.increment_set_count()
future_set_count = self.helpers.get_set_count()
if current_set_count == future_set_count-1:
out = f"[AMDSMI_STATUS_NOT_SUPPORTED] Unable to set compute partition to {user_requested_partition_args}"
self.logger.store_output(args.gpu, 'accelerator_partition', out)
elif e.get_error_code() == amdsmi_interface.amdsmi_wrapper.AMDSMI_STATUS_SETTING_UNAVAILABLE:
print(f"\n{attempted_to_set}\n"
f"\n[AMDSMI_STATUS_SETTING_UNAVAILABLE] Please check amd-smi partition --memory --accelerator for available profiles.\n"
"Users may need to switch memory partition to another mode in order to enable the desired accelerator partition.\n")
raise ValueError(f"[AMDSMI_STATUS_SETTING_UNAVAILABLE] Unable to set accelerator partition to {args.compute_partition} on {gpu_string}") from e
else:
raise ValueError(f"Unable to set accelerator partition to {args.compute_partition} on {gpu_string}") from e
self.logger.print_output()
self.logger.clear_multiple_devices_output()
return
if args.memory_partition:
####################################################################
# Get current and available memory partition modes #
# Info used if AMDSMI_STATUS_INVAL is caught & to set progress bar #
####################################################################
self.helpers.increment_set_count()
set_count = self.helpers.get_set_count()
if set_count == 1: # only show reload warning on 1st set
self.helpers.confirm_changing_memory_partition_gpu_reload_warning()
try:
memory_dict = {'caps': "N/A", 'current': "N/A"}
memory_partition_config = amdsmi_interface.amdsmi_get_gpu_memory_partition_config(args.gpu)
memory_dict['caps'] = str(memory_partition_config['partition_caps']).replace("]", "").replace("[", "").replace("\'", "").replace(" ", "")
memory_dict['current'] = memory_partition_config['mp_mode']
except amdsmi_exception.AmdSmiLibraryException as e:
logging.debug("Failed to get current memory partition for GPU %s | %s", gpu_id, e.get_error_info())
try:
memory_partition = amdsmi_interface.AmdSmiMemoryPartitionType[args.memory_partition]
amdsmi_interface.amdsmi_set_gpu_memory_partition(args.gpu, memory_partition)
out = f"Successfully set memory partition to {args.memory_partition}, reload driver when ready"
except amdsmi_exception.AmdSmiLibraryException as e:
out = f"[{e.get_error_info(detailed=False)}] Unable to set memory partition to {args.memory_partition}"
if e.get_error_code() == amdsmi_interface.amdsmi_wrapper.AMDSMI_STATUS_NO_PERM:
out = f"[AMDSMI_STATUS_NO_PERM] Command requires elevation"
self.logger.store_output(args.gpu, 'memory_partition', out)
self.logger.print_output()
self.logger.clear_multiple_devices_output()
raise PermissionError('Command requires elevation') from e
elif e.get_error_code() == amdsmi_interface.amdsmi_wrapper.AMDSMI_STATUS_INVAL:
print(f"Valid Memory partition Modes: {memory_dict['caps']}\n")
self.logger.store_output(args.gpu, 'memory_partition', out)
self.logger.print_output()
self.logger.clear_multiple_devices_output()
return
else:
self.logger.store_output(args.gpu, 'memory_partition', out)
self.logger.print_output()
self.logger.clear_multiple_devices_output()
return
self.logger.store_output(args.gpu, 'memory_partition', out)
self.logger.print_output()
self.logger.clear_multiple_devices_output()
return
if isinstance(args.soc_pstate, int):
try:
amdsmi_interface.amdsmi_set_soc_pstate(args.gpu, args.soc_pstate)
except amdsmi_exception.AmdSmiLibraryException as e:
if e.get_error_code() == amdsmi_interface.amdsmi_wrapper.AMDSMI_STATUS_NO_PERM:
raise PermissionError('Command requires elevation') from e
if e.get_error_code() == amdsmi_interface.amdsmi_wrapper.AMDSMI_STATUS_INVAL:
soc_pstate_info = amdsmi_interface.amdsmi_get_soc_pstate(args.gpu)
policy_string = "N/A"
# Check if 'policies' key exists before accessing it
if 'policies' in soc_pstate_info and soc_pstate_info['policies']:
policy_string = ""
for policy in soc_pstate_info['policies']:
policy_string += f"{policy['policy_id']}: {policy['policy_description']}, "
policy_string = policy_string.rstrip(", ") # Remove trailing comma and space
print(f"Valid SOC P-State Policies: [{policy_string}]\n")
self.logger.store_output(args.gpu, 'socpstate', f"[{e.get_error_info(detailed=False)}] Unable to set soc pstate dpm policy to {args.soc_pstate}")
self.logger.print_output()
self.logger.clear_multiple_devices_output()
return
self.logger.store_output(args.gpu, 'socpstate', f"Successfully set soc pstate dpm policy to {args.soc_pstate}")
self.logger.print_output()
self.logger.clear_multiple_devices_output()
return
if isinstance(args.xgmi_plpd, int):
try:
amdsmi_interface.amdsmi_set_xgmi_plpd(args.gpu, args.xgmi_plpd)
except amdsmi_exception.AmdSmiLibraryException as e:
if e.get_error_code() == amdsmi_interface.amdsmi_wrapper.AMDSMI_STATUS_NO_PERM:
raise PermissionError('Command requires elevation') from e
if e.get_error_code() == amdsmi_interface.amdsmi_wrapper.AMDSMI_STATUS_INVAL:
xgmi_plpd_info = amdsmi_interface.amdsmi_get_xgmi_plpd(args.gpu)
policy_string = "N/A"
# Check if 'policies' key exists before accessing it
if 'policies' in xgmi_plpd_info and xgmi_plpd_info['policies']:
policy_string = ""
for policy in xgmi_plpd_info['policies']:
policy_string += f"{policy['policy_id']}: {policy['policy_description']}, "
policy_string = policy_string.rstrip(", ") # Remove trailing comma and space
print(f"Valid XGMI PLPD Policies: [{policy_string}]\n")
self.logger.store_output(args.gpu, 'xgmiplpd', f"[{e.get_error_info(detailed=False)}] Unable to set XGMI per-link power down policy to {args.xgmi_plpd}")
self.logger.print_output()
self.logger.clear_multiple_devices_output()
return
self.logger.store_output(args.gpu, 'xgmiplpd', f"Successfully set XGMI per-link power down policy to {args.xgmi_plpd}")
self.logger.print_output()
self.logger.clear_multiple_devices_output()
return
if isinstance(args.clk_level, tuple):
clk_type = args.clk_level.clk_type
perf_levels = args.clk_level.perf_levels
perf_levels_str = str(perf_levels).strip('[]').replace(" ", "")
smi_clk_type_mapping = {
"sclk": amdsmi_interface.AmdSmiClkType.SYS,
"mclk": amdsmi_interface.AmdSmiClkType.MEM,
"pcie": amdsmi_interface.AmdSmiClkType.PCIE,
"fclk": amdsmi_interface.AmdSmiClkType.DF,
"socclk": amdsmi_interface.AmdSmiClkType.SOC
}
results_clk_lvl = {'perf_level': f"Unable to set performance level to MANUAL",
'get_clock_freq': f"Unable to retrieve {clk_type} frequency levels",
'set_clock': f"Unable to set {clk_type} perf level(s) to {perf_levels_str}"}
if clk_type not in smi_clk_type_mapping:
raise ValueError(f"Invalid clock type {clk_type}. Valid options are: {', '.join(smi_clk_type_mapping.keys())}")
# Set perf level to manual if not already set
try:
amdsmi_interface.amdsmi_set_gpu_perf_level(args.gpu, amdsmi_interface.AmdSmiDevPerfLevel.MANUAL)
results_clk_lvl['perf_level'] = f"Successfully set performance level to MANUAL"
except amdsmi_exception.AmdSmiLibraryException as e:
if e.get_error_code() == amdsmi_interface.amdsmi_wrapper.AMDSMI_STATUS_NO_PERM:
raise PermissionError('Command requires elevation') from e
results_clk_lvl['perf_level'] = f"[{e.get_error_info(detailed=False)}] Unable to set performance level to MANUAL"
self.logger.store_output(args.gpu, 'clk_level', results_clk_lvl)
self.logger.print_output()
self.logger.clear_multiple_devices_output()
return
if clk_type.lower() == "pcie":
# Get PCIe bandwidth levels
try:
pcie_bandwidth_levels = amdsmi_interface.amdsmi_get_gpu_pci_bandwidth(args.gpu)
num_supported = pcie_bandwidth_levels['transfer_rate']['num_supported']
results_clk_lvl['get_clock_freq'] = f"Successfully retrieved {clk_type} frequency levels"
except amdsmi_exception.AmdSmiLibraryException as e:
results_clk_lvl['get_clock_freq'] = f"[{e.get_error_info(detailed=False)}] Unable to retrieve {clk_type} frequency levels"
self.logger.store_output(args.gpu, 'clk_level', results_clk_lvl)
self.logger.print_output()
self.logger.clear_multiple_devices_output()
return
else:
# Get clock frequency levels
try:
frequencies = amdsmi_interface.amdsmi_get_clk_freq(args.gpu, smi_clk_type_mapping[clk_type])
num_supported = frequencies['num_supported']
results_clk_lvl['get_clock_freq'] = f"Successfully retrieved {clk_type} frequency levels"
except amdsmi_exception.AmdSmiLibraryException as e:
results_clk_lvl['get_clock_freq'] = f"[{e.get_error_info(detailed=False)}] Unable to retrieve {clk_type} frequency levels"
self.logger.store_output(args.gpu, 'clk_level', results_clk_lvl)
self.logger.print_output()
self.logger.clear_multiple_devices_output()
return
# Validate bandwidth bitmask
freq_bitmask = 0
invalid_levels = []
for level in perf_levels:
if level < num_supported:
freq_bitmask |= (1 << level)
else:
invalid_levels.append(level)
if invalid_levels:
# Handle/report invalid levels
invalid_levels_str = str(invalid_levels).strip('[]').replace(" ", "")
valid_levels_str = f"Valid levels for {clk_type}: 0"
if num_supported > 1:
valid_levels_str = f"Valid levels for {clk_type}: 0-{num_supported-1}"
print(f"\n{valid_levels_str}\n")
results_clk_lvl['set_clock'] = f"Invalid level(s) {invalid_levels_str} are not within the range of supported levels for {clk_type}"
self.logger.store_output(args.gpu, 'clk_level', results_clk_lvl)
self.logger.print_output()
self.logger.clear_multiple_devices_output()
return
else:
# Proceed with freq_bitmask
pass
if clk_type.lower() == "pcie":
try:
amdsmi_interface.amdsmi_set_gpu_pci_bandwidth(args.gpu, freq_bitmask)
results_clk_lvl['set_clock'] = f"Successfully set {clk_type} perf level(s) to {perf_levels_str}"
except amdsmi_exception.AmdSmiLibraryException as e:
if e.get_error_code() == amdsmi_interface.amdsmi_wrapper.AMDSMI_STATUS_NO_PERM:
raise PermissionError('Command requires elevation') from e
results_clk_lvl['set_clock'] = f"[{e.get_error_info(detailed=False)}] Unable to set {clk_type} perf level(s) to {perf_levels_str}"
self.logger.store_output(args.gpu, 'clk_level', results_clk_lvl)
self.logger.print_output()
self.logger.clear_multiple_devices_output()
return
else:
# For non-pcie clocks
if clk_type in self.convert_clock_type:
clk_type_conversion = self.convert_clock_type[clk_type]
else:
clk_type_conversion = "N/A"
try:
amdsmi_interface.amdsmi_set_clk_freq(args.gpu, clk_type, freq_bitmask)
results_clk_lvl['set_clock'] = f"Successfully set {clk_type} perf level(s) to {perf_levels_str}"
except amdsmi_exception.AmdSmiLibraryException as e:
if e.get_error_code() == amdsmi_interface.amdsmi_wrapper.AMDSMI_STATUS_NO_PERM:
raise PermissionError('Command requires elevation') from e
results_clk_lvl['set_clock'] = f"[{e.get_error_info(detailed=False)}] Unable to set {clk_type} perf level(s) to {perf_levels_str}"
self.logger.store_output(args.gpu, 'clk_level', results_clk_lvl)
self.logger.print_output()
self.logger.clear_multiple_devices_output()
return
self.logger.store_output(args.gpu, 'clk_level', results_clk_lvl)
self.logger.print_output()
self.logger.clear_multiple_devices_output()
return
if isinstance(args.ptl_status, int):
status_string = "Enabled" if args.ptl_status else "Disabled"
result = f"Requested PTL status to {status_string}" # This should not print out
try: # Due to driver requirements, do NOT check current state. Set state regardless of current state.
amdsmi_interface.amdsmi_set_gpu_ptl_state(args.gpu, args.ptl_status)
result = f"Successfully set PTL state to {status_string}"
except amdsmi_exception.AmdSmiLibraryException as e:
if e.get_error_code() == amdsmi_interface.amdsmi_wrapper.AMDSMI_STATUS_NO_PERM:
raise PermissionError('Command requires elevation') from e
self.logger.store_output(args.gpu, 'ptlstatus', f"[{e.get_error_info(detailed=False)}] Unable to set ptl status to {args.ptl_status}")
self.logger.print_output()
self.logger.clear_multiple_devices_output()
return
self.logger.store_output(args.gpu, 'ptlstatus', result)
self.logger.print_output()
self.logger.clear_multiple_devices_output()
return
if isinstance(args.ptl_format, tuple):
requested_fmt1_enum, requested_fmt2_enum = args.ptl_format
requested_str = f"{requested_fmt1_enum.name},{requested_fmt2_enum.name}"
result = f"Requested PTL status to {requested_str}" # This should not print out
try:
# Get current formats as ints
cur1_code, cur2_code = amdsmi_interface.amdsmi_get_gpu_ptl_formats(args.gpu)
cur1_enum = amdsmi_interface.AmdSmiPtlData(cur1_code)
cur2_enum = amdsmi_interface.AmdSmiPtlData(cur2_code)
current_str = f"{cur1_enum.name},{cur2_enum.name}"
if (cur1_enum, cur2_enum) == (requested_fmt1_enum, requested_fmt2_enum):
result = f"PTL format is already {current_str}"
else:
amdsmi_interface.amdsmi_set_gpu_ptl_formats(args.gpu, requested_fmt1_enum, requested_fmt2_enum)
result = f"Successfully set PTL format to {requested_str}"
except amdsmi_exception.AmdSmiLibraryException as e:
if e.get_error_code() == amdsmi_interface.amdsmi_wrapper.AMDSMI_STATUS_NO_PERM:
raise PermissionError('Command requires elevation') from e
self.logger.store_output(args.gpu, 'ptlformat', f"[{e.get_error_info(detailed=False)}] Unable to set PTL format to {requested_str}")
self.logger.print_output()
self.logger.clear_multiple_devices_output()
return
self.logger.store_output(args.gpu, 'ptlformat', result)
self.logger.print_output()
self.logger.clear_multiple_devices_output()
return
# Universal args
if isinstance(args.power_cap, tuple):
pwr_type = args.power_cap.pwr_type
requested_power_cap = args.power_cap.watts
# If pwr_type is None, default to ppt0 (legacy behavior)
if pwr_type is None:
pwr_type = "ppt0"
pwr_type_as_int = 0
else:
pwr_type_as_int = 0 if pwr_type == "ppt0" else 1
# Set the power cap for the specified sensor
pwr_type_upper = pwr_type.upper()
result = self.helpers.validate_and_set_power_cap(
args.gpu, pwr_type_as_int, pwr_type_upper, requested_power_cap, self.logger)
self.logger.store_output(args.gpu, 'powercap', result)
if multiple_devices:
self.logger.store_multiple_device_output()
return # Skip printing when there are multiple devices
self.logger.print_output()
self.logger.clear_multiple_devices_output()
return
if isinstance(args.clk_limit, tuple):
clk_type = args.clk_limit.clk_type
lim_type = args.clk_limit.lim_type
val = args.clk_limit.val
val_changed = True # Assume Clock limit value is changed
# Validate the value against the extremum
try:
# Parser only allows two options sclk or mclk
if clk_type == "sclk":
amdsmi_clk_type = amdsmi_interface.AmdSmiClkType.GFX
elif clk_type == "mclk":
amdsmi_clk_type = amdsmi_interface.AmdSmiClkType.MEM
else:
print(f"Valid clock types are: sclk, mclk\n")
self.logger.store_output(args.gpu, 'clk_limit', f"Invalid clock type {args.clk_limit.clk_type}")
self.logger.print_output()
self.logger.clear_multiple_devices_output()
return
clk_tuple = amdsmi_interface.amdsmi_get_clock_info(args.gpu, amdsmi_clk_type)
if lim_type == "min":
amdsmi_lim_type = amdsmi_interface.AmdSmiClkLimitType.MIN
if val > clk_tuple['max_clk']:
self.logger.store_output(args.gpu, 'clk_limit', f"Cannot set {args.clk_limit.clk_type} min value greater than max ({clk_tuple['max_clk']}MHz)")
self.logger.print_output()
self.logger.clear_multiple_devices_output()
return
if val == clk_tuple['min_clk']:
val_changed = False # Clock limit value did not changed
elif lim_type == "max":
amdsmi_lim_type = amdsmi_interface.AmdSmiClkLimitType.MAX
if val < clk_tuple['min_clk']:
self.logger.store_output(args.gpu, 'clk_limit', f"Cannot set {args.clk_limit.clk_type} max value less than min ({clk_tuple['min_clk']}MHz)")
self.logger.print_output()
self.logger.clear_multiple_devices_output()
return
if val == clk_tuple['max_clk']:
val_changed = False # Clock limit value did not changed
except amdsmi_exception.AmdSmiLibraryException as e:
if e.get_error_code() == amdsmi_interface.amdsmi_wrapper.AMDSMI_STATUS_NOT_SUPPORTED and lim_type == "min" and clk_type == "mclk":
logging.debug("Setting mclk min is not supported")
self.logger.store_output(args.gpu, 'clk_limit', f"Setting mclk min is not supported")
else:
logging.debug("Failed to get clock extremum info for gpu %s | %s", gpu_id, e.get_error_info())
self.logger.store_output(args.gpu, 'clk_limit', f"[{e.get_error_info(detailed=False)}] Unable to change {args.clk_limit.lim_type} of {args.clk_limit.clk_type} to {args.clk_limit.val}MHz")
self.logger.print_output()
self.logger.clear_multiple_devices_output()
return
# Set the value
try:
if val_changed:
amdsmi_interface.amdsmi_set_gpu_clk_limit(args.gpu, clk_type, lim_type, val)
except amdsmi_exception.AmdSmiLibraryException as e:
if e.get_error_code() == amdsmi_interface.amdsmi_wrapper.AMDSMI_STATUS_NO_PERM:
raise PermissionError('Command requires elevation') from e
elif e.get_error_code() == amdsmi_interface.amdsmi_wrapper.AMDSMI_STATUS_NOT_SUPPORTED and lim_type == "min" and clk_type == "mclk":
logging.debug("Setting mclk min is not supported")
self.logger.store_output(args.gpu, 'clk_limit', f"Setting mclk min is not supported")
else:
self.logger.store_output(args.gpu, 'clk_limit', f"[{e.get_error_info(detailed=False)}] Unable to set {args.clk_limit.lim_type} of {args.clk_limit.clk_type} to {args.clk_limit.val}MHz")
self.logger.print_output()
self.logger.clear_multiple_devices_output()
return
if val_changed:
self.logger.store_output(args.gpu, 'clk_limit', f"Successfully changed {args.clk_limit.lim_type} of {args.clk_limit.clk_type} to {args.clk_limit.val}MHz")
else:
self.logger.store_output(args.gpu, 'clk_limit', f"Clock limit is already set to {args.clk_limit.val}MHz")
self.logger.print_output()
self.logger.clear_multiple_devices_output()
return
if isinstance(args.process_isolation, int):
status_string = "Enabled" if args.process_isolation else "Disabled"
result = f"Requested process isolation to {status_string}" # This should not print out
try:
current_status = amdsmi_interface.amdsmi_get_gpu_process_isolation(args.gpu)
if current_status == args.process_isolation:
result = f"Process isolation is already {status_string}"
else:
amdsmi_interface.amdsmi_set_gpu_process_isolation(args.gpu, args.process_isolation)
result = f"Successfully set process isolation to {status_string}"
except amdsmi_exception.AmdSmiLibraryException as e:
if e.get_error_code() == amdsmi_interface.amdsmi_wrapper.AMDSMI_STATUS_NO_PERM:
raise PermissionError('Command requires elevation') from e
self.logger.store_output(args.gpu, 'process_isolation', f"[{e.get_error_info(detailed=False)}] Unable to set process isolation to {status_string}")
self.logger.print_output()
self.logger.clear_multiple_devices_output()
return
self.logger.store_output(args.gpu, 'process_isolation', result)
self.logger.print_output()
self.logger.clear_multiple_devices_output()
return
def set_value(self, args, multiple_devices=False, gpu=None, fan=None, perf_level=None,
profile=None, perf_determinism=None, compute_partition=None,
memory_partition=None, power_cap=None,
cpu=None, cpu_pwr_limit=None, cpu_xgmi_link_width=None, cpu_lclk_dpm_level=None,
cpu_pwr_eff_mode=None, cpu_gmi3_link_width=None, cpu_pcie_link_rate=None,
cpu_df_pstate_range=None, cpu_enable_apb=None, cpu_disable_apb=None,
soc_boost_limit=None, core=None, core_boost_limit=None, soc_pstate=None, xgmi_plpd=None,
process_isolation=None, clk_limit=None, clk_level=None, cpu_dfcstate_ctrl=None,
cpu_railisofreq_policy=None, ptl_status=None, ptl_format=None):
"""Issue reset commands to target gpu(s)
Args:
args (Namespace): Namespace containing the parsed CLI args
multiple_devices (bool, optional): True if checking for multiple devices. Defaults to False.
gpu (device_handle, optional): device_handle for target device. Defaults to None.
fan (int, optional): Value override for args.fan. Defaults to None.
perf_level (amdsmi_interface.AmdSmiDevPerfLevel, optional): Value override for args.perf_level. Defaults to None.
profile (bool, optional): Value override for args.profile. Defaults to None.
perf_determinism (int, optional): Value override for args.perf_determinism. Defaults to None.
compute_partition (amdsmi_interface.AmdSmiComputePartitionType, optional): Value override for args.compute_partition. Defaults to None.
memory_partition (amdsmi_interface.AmdSmiMemoryPartitionType, optional): Value override for args.memory_partition. Defaults to None.
power_cap (int, optional): Value override for args.power_cap. Defaults to None.
cpu (cpu_handle, optional): device_handle for target device. Defaults to None.
cpu_pwr_limit (int, optional): Value override for args.cpu_pwr_limit. Defaults to None.
cpu_xgmi_link_width (List[int], optional): Value override for args.cpu_xgmi_link_width. Defaults to None.
cpu_lclk_dpm_level (List[int], optional): Value override for args.cpu_lclk_dpm_level. Defaults to None.
cpu_pwr_eff_mode (int, optional): Value override for args.cpu_pwr_eff_mode. Defaults to None.
cpu_gmi3_link_width (List[int], optional): Value override for args.cpu_gmi3_link_width. Defaults to None.
cpu_pcie_link_rate (int, optional): Value override for args.cpu_pcie_link_rate. Defaults to None.
cpu_df_pstate_range (List[int], optional): Value override for args.cpu_df_pstate_range. Defaults to None.
cpu_enable_apb (bool, optional): Value override for args.cpu_enable_apb. Defaults to None.
cpu_disable_apb (int, optional): Value override for args.cpu_disable_apb. Defaults to None.
soc_boost_limit (int, optional): Value override for args.soc_boost_limit. Defaults to None.
cpu_dfcstate_ctrl (int, optional): Value override for args.cpu_dfcstate_ctrl. Defaults to None.
cpu_railisofreq_policy (int, optional): Value override for args.cpu_railisofreq_policy. Defaults to None.
core (device_handle, optional): device_handle for target core. Defaults to None.
core_boost_limit (int, optional): Value override for args.core_boost_limit. Defaults to None
soc_pstate (int, optional): Value override for args.soc_pstate. Defaults to None.
xgmi_plpd (int, optional): Value override for args.xgmi_plpd. Defaults to None.
process_isolation (int, optional): Value override for args.process_isolation. Defaults to None.
Raises:
ValueError: Value error if no gpu value is provided
IndexError: Index error if gpu list is empty
Return:
Nothing
"""
# These are the only args checked at this point, the other args will be passed
# in through the applicable function set_gpu, set_cpu, or set_core function
if gpu:
args.gpu = gpu
if cpu:
args.cpu = cpu
if core:
args.core = core
# Check if a GPU argument has been set
gpu_args_enabled = False
gpu_attributes = ["fan", "perf_level", "profile", "perf_determinism", "compute_partition",
"memory_partition", "power_cap", "soc_pstate", "xgmi_plpd",
"process_isolation", "clk_limit", "clk_level", "ptl_status", "ptl_format"]
for attr in gpu_attributes:
if hasattr(args, attr):
if getattr(args, attr) is not None:
gpu_args_enabled = True
break
# Check if a CPU argument has been set
cpu_args_enabled = False
cpu_attributes = ["cpu_pwr_limit", "cpu_xgmi_link_width", "cpu_lclk_dpm_level", "cpu_pwr_eff_mode",
"cpu_gmi3_link_width", "cpu_pcie_link_rate", "cpu_df_pstate_range",
"cpu_enable_apb", "cpu_disable_apb", "soc_boost_limit",
"cpu_dfcstate_ctrl", "cpu_railisofreq_policy"]
for attr in cpu_attributes:
if hasattr(args, attr):
if getattr(args, attr) not in [None, False]:
cpu_args_enabled = True
break
# Check if a Core argument has been set
core_args_enabled = False
core_attributes = ["core_boost_limit"]
for attr in core_attributes:
if hasattr(args, attr):
if getattr(args, attr) is not None:
core_args_enabled = True
break
# Error if no subcommand args are passed
if self.helpers.is_baremetal():
is_gpu_set = False
is_cpu_set = False
is_core_set = False
try:
is_gpu_set = any([
args.gpu is not None,
args.fan is not None,
args.perf_level is not None,
args.profile is not None,
args.perf_determinism is not None,
args.compute_partition is not None,
args.memory_partition is not None,
args.power_cap is not None,
args.soc_pstate is not None,
args.xgmi_plpd is not None,
args.clk_limit is not None,
args.clk_level is not None,
args.ptl_status is not None,
args.ptl_format is not None,
args.process_isolation is not None
])
except AttributeError:
# If attribute error for gpu, then we could be another subcommand
pass
try:
is_cpu_set = any([
args.cpu is not None,
args.cpu_pwr_limit is not None,
args.cpu_xgmi_link_width is not None,
args.cpu_lclk_dpm_level is not None,
args.cpu_pwr_eff_mode is not None,
args.cpu_gmi3_link_width is not None,
args.cpu_pcie_link_rate is not None,
args.cpu_df_pstate_range is not None,
args.cpu_enable_apb,
args.cpu_disable_apb is not None,
args.soc_boost_limit is not None,
args.cpu_dfcstate_ctrl is not None,
args.cpu_railisofreq_policy is not None
])
except AttributeError:
# If attribute error for cpu, then we could be another subcommand
pass
try:
if args.core_boost_limit:
is_core_set = True
except AttributeError:
# If attribute error for core, then we could be another subcommand
pass
if not (is_gpu_set or is_cpu_set or is_core_set):
# if neither GPU / CPU / or Core args are provided, then raise error message
command = " ".join(sys.argv[1:])
raise AmdSmiRequiredCommandException(command, self.logger.format)
else:
if not any([args.process_isolation is not None, args.clk_limit is not None, args.power_cap is not None]):
command = " ".join(sys.argv[1:])
raise AmdSmiRequiredCommandException(command, self.logger.format)
# Only allow one device's arguments to be set at a time
if not any([gpu_args_enabled, cpu_args_enabled, core_args_enabled]):
raise ValueError('No GPU, CPU, or CORE arguments provided, specific arguments are needed')
elif all([gpu_args_enabled, cpu_args_enabled, core_args_enabled]):
raise ValueError('Cannot set GPU, CPU, and CORE arguments at the same time')
elif not (gpu_args_enabled ^ cpu_args_enabled ^ core_args_enabled):
raise ValueError('Cannot set GPU, CPU, or CORE arguments at the same time')
if self.helpers.is_amdgpu_initialized() and gpu_args_enabled:
if args.gpu == None:
args.gpu = self.device_handles
if self.helpers.is_amd_hsmp_initialized() and cpu_args_enabled:
if args.cpu == None:
args.cpu = self.cpu_handles
if self.helpers.is_amd_hsmp_initialized() and core_args_enabled:
if args.core == None:
args.core = self.core_handles
# Handle CPU and GPU intialization cases
if self.helpers.is_amd_hsmp_initialized() and self.helpers.is_amdgpu_initialized():
# Print out all CPU and all GPU static info only if no device was specified.
# If a GPU or CPU argument is provided only print out the specified device.
if args.cpu == None and args.gpu == None and args.core == None:
raise ValueError('No GPU, CPU, or CORE provided, specific target(s) are needed')
if args.cpu:
self.set_cpu(args, multiple_devices, cpu, cpu_pwr_limit,
cpu_xgmi_link_width, cpu_lclk_dpm_level, cpu_pwr_eff_mode,
cpu_gmi3_link_width, cpu_pcie_link_rate, cpu_df_pstate_range,
cpu_enable_apb, cpu_disable_apb, soc_boost_limit,
cpu_dfcstate_ctrl, cpu_railisofreq_policy)
if args.core:
self.logger.output = {}
self.logger.clear_multiple_devices_output()
self.set_core(args, multiple_devices, core, core_boost_limit)
if args.gpu:
self.logger.output = {}
self.logger.clear_multiple_devices_output()
self.set_gpu(args, multiple_devices, gpu, fan, perf_level,
profile, perf_determinism, compute_partition,
memory_partition, power_cap, soc_pstate, xgmi_plpd,
process_isolation, clk_limit, clk_level, ptl_status, ptl_format)
elif self.helpers.is_amd_hsmp_initialized(): # Only CPU is initialized
if args.cpu == None and args.core == None:
raise ValueError('No CPU or CORE provided, specific target(s) are needed')
if args.cpu:
self.set_cpu(args, multiple_devices, cpu, cpu_pwr_limit,
cpu_xgmi_link_width, cpu_lclk_dpm_level, cpu_pwr_eff_mode,
cpu_gmi3_link_width, cpu_pcie_link_rate, cpu_df_pstate_range,
cpu_enable_apb, cpu_disable_apb, soc_boost_limit,
cpu_dfcstate_ctrl, cpu_railisofreq_policy)
if args.core:
self.logger.output = {}
self.logger.clear_multiple_devices_output()
self.set_core(args, multiple_devices, core, core_boost_limit)
elif self.helpers.is_amdgpu_initialized(): # Only GPU is initialized
if args.gpu == None:
args.gpu = self.device_handles
self.logger.clear_multiple_devices_output()
self.set_gpu(args, multiple_devices, gpu, fan, perf_level,
profile, perf_determinism, compute_partition,
memory_partition, power_cap, soc_pstate, xgmi_plpd,
process_isolation, clk_limit, clk_level, ptl_status, ptl_format)
def reset(self, args, multiple_devices=False, gpu=None, gpureset=None,
clocks=None, fans=None, profile=None, xgmierr=None, perf_determinism=None,
power_cap=None, reload_driver=None, clean_local_data=None):
"""Issue reset commands to target gpu(s)
Args:
args (Namespace): Namespace containing the parsed CLI args
multiple_devices (bool, optional): True if checking for multiple devices. Defaults to False.
gpu (device_handle, optional): device_handle for target device. Defaults to None.
gpureset (bool, optional): Value override for args.gpureset. Defaults to None.
clocks (bool, optional): Value override for args.clocks. Defaults to None.
fans (bool, optional): Value override for args.fans. Defaults to None.
profile (bool, optional): Value override for args.profile. Defaults to None.
xgmierr (bool, optional): Value override for args.xgmierr. Defaults to None.
perf_determinism (bool, optional): Value override for args.perf_determinism. Defaults to None.
power_cap (bool, optional): Value override for args.power_cap. Defaults to None.
clean_local_data (bool, optional): Value override for args.run_cleaner_shader. Defaults to None.
Raises:
ValueError: Value error if no gpu value is provided
IndexError: Index error if gpu list is empty
Return:
Nothing
"""
# Set args.* to passed in arguments
if gpu:
args.gpu = gpu
if gpureset:
args.gpureset = gpureset
if clocks:
args.clocks = clocks
if fans:
args.fans = fans
if profile:
args.profile = profile
if xgmierr:
args.xgmierr = xgmierr
if perf_determinism:
args.perf_determinism = perf_determinism
if power_cap:
args.power_cap = power_cap
if reload_driver:
args.reload_driver = reload_driver
if clean_local_data:
args.clean_local_data = clean_local_data
# Handle No GPU passed
if args.gpu == None:
args.gpu = self.device_handles
if not self.group_check_printed:
self.helpers.check_required_groups(check_render=True, check_video=False)
self.group_check_printed = True
# Mode-1 gpureset is hive-wide.
# Group GPUs by hive and reset each hive only once.
gpus_to_reset = []
if args.gpureset and isinstance(args.gpu, list) and len(args.gpu) > 1:
# Group GPUs by their XGMI hive ID.
# If GPU not in a hive or no hive info, reset individually.
hive_to_gpus = {}
gpus_without_hive = []
for gpu in args.gpu:
try:
xgmi_info = amdsmi_interface.amdsmi_get_xgmi_info(gpu)
if isinstance(xgmi_info, dict):
hive_id = xgmi_info.get('xgmi_hive_id', None)
if hive_id is not None and hive_id != 0:
if hive_id not in hive_to_gpus:
hive_to_gpus[hive_id] = []
hive_to_gpus[hive_id].append(gpu)
else:
gpus_without_hive.append(gpu)
else:
gpus_without_hive.append(gpu)
except:
gpus_without_hive.append(gpu)
# For each hive, reset using the first GPU (resets entire hive)
for hive_id, gpu_list in hive_to_gpus.items():
gpus_to_reset.append(gpu_list[0])
# Add all non-hive GPUs to reset individually
gpus_to_reset.extend(gpus_without_hive)
# Update args.gpu to only the GPUs to reset
if gpus_to_reset:
args.gpu = gpus_to_reset
# Handle multiple GPUs
handled_multiple_gpus, device_handle = self.helpers.handle_gpus(args, self.logger, self.reset)
if handled_multiple_gpus:
return # This function is recursive
args.gpu = device_handle
# Get gpu_id for logging
gpu_id = self.helpers.get_gpu_id_from_device_handle(args.gpu)
# Error if no subcommand args are passed
if self.helpers.is_baremetal():
if not any([args.gpureset, args.clocks, args.fans, args.profile, args.xgmierr, \
args.perf_determinism, args.power_cap, args.reload_driver, \
args.clean_local_data]):
command = " ".join(sys.argv[1:])
raise AmdSmiRequiredCommandException(command, self.logger.format)
else:
if not any([args.clean_local_data, args.reload_driver]):
command = " ".join(sys.argv[1:])
raise AmdSmiRequiredCommandException(command, self.logger.format)
#######################
# BM commands - START #
#######################
if self.helpers.is_baremetal():
if args.gpureset:
if self.helpers.is_amd_device(args.gpu):
try:
amdsmi_interface.amdsmi_reset_gpu(args.gpu)
result = 'Successfully reset GPU'
except amdsmi_exception.AmdSmiLibraryException as e:
if e.get_error_code() == amdsmi_interface.amdsmi_wrapper.AMDSMI_STATUS_NO_PERM:
raise PermissionError('Command requires elevation') from e
result = f"[{e.get_error_info(detailed=False)}] Unable to reset GPU"
self.logger.store_output(args.gpu, 'gpu_reset', result)
self.logger.print_output()
self.logger.clear_multiple_devices_output()
return
else:
result = 'Unable to reset non-amd GPU'
self.logger.store_output(args.gpu, 'gpu_reset', result)
self.logger.print_output()
self.logger.clear_multiple_devices_output()
return
if args.clocks:
reset_clocks_results = {'overdrive': '',
'clocks': '',
'performance': ''}
try:
amdsmi_interface.amdsmi_set_gpu_overdrive_level(args.gpu, 0)
reset_clocks_results['overdrive'] = 'Overdrive set to 0'
except amdsmi_exception.AmdSmiLibraryException as e:
if e.get_error_code() == amdsmi_interface.amdsmi_wrapper.AMDSMI_STATUS_NO_PERM:
raise PermissionError('Command requires elevation') from e
logging.debug("Failed to reset overdrive on gpu %s | %s", gpu_id, e.get_error_info())
reset_clocks_results['overdrive'] = f"[{e.get_error_info(detailed=False)}] Unable to reset overdrive to 0"
# continue to reset clocks and performance level
try:
level_auto = amdsmi_interface.AmdSmiDevPerfLevel.AUTO
amdsmi_interface.amdsmi_set_gpu_perf_level(args.gpu, level_auto)
reset_clocks_results['clocks'] = 'Successfully reset performance level to auto'
except amdsmi_exception.AmdSmiLibraryException as e:
if e.get_error_code() == amdsmi_interface.amdsmi_wrapper.AMDSMI_STATUS_NO_PERM:
raise PermissionError('Command requires elevation') from e
reset_clocks_results['clocks'] = f"[{e.get_error_info(detailed=False)}] Unable to reset performance level to auto"
logging.debug("Failed to reset perf level on gpu %s | %s", gpu_id, e.get_error_info())
try:
#TODO: Check why this is called twice?
level_auto = amdsmi_interface.AmdSmiDevPerfLevel.AUTO
amdsmi_interface.amdsmi_set_gpu_perf_level(args.gpu, level_auto)
reset_clocks_results['performance'] = 'Successfully reset performance level to auto'
except amdsmi_exception.AmdSmiLibraryException as e:
if e.get_error_code() == amdsmi_interface.amdsmi_wrapper.AMDSMI_STATUS_NO_PERM:
raise PermissionError('Command requires elevation') from e
reset_clocks_results['performance'] = f"[{e.get_error_info(detailed=False)}] Unable to reset performance level to auto"
logging.debug("Failed to reset perf level on gpu %s | %s", gpu_id, e.get_error_info())
self.logger.store_output(args.gpu, 'reset_clocks', reset_clocks_results)
self.logger.print_output()
self.logger.clear_multiple_devices_output()
return
if args.fans:
try:
amdsmi_interface.amdsmi_reset_gpu_fan(args.gpu, 0)
result = 'Successfully reset fan speed to driver control'
except amdsmi_exception.AmdSmiLibraryException as e:
if e.get_error_code() == amdsmi_interface.amdsmi_wrapper.AMDSMI_STATUS_NO_PERM:
raise PermissionError('Command requires elevation') from e
result = f"[{e.get_error_info(detailed=False)}] Unable to reset fan speed to driver control"
logging.debug("Failed to reset fans on gpu %s | %s", gpu_id, e.get_error_info())
self.logger.store_output(args.gpu, 'reset_fans', result)
self.logger.print_output()
self.logger.clear_multiple_devices_output()
return
self.logger.store_output(args.gpu, 'reset_fans', result)
self.logger.print_output()
self.logger.clear_multiple_devices_output()
return
if args.profile:
reset_profile_results = {'power_profile' : 'N/A'}
try:
power_profile_mask = amdsmi_interface.AmdSmiPowerProfilePresetMasks.BOOTUP_DEFAULT
amdsmi_interface.amdsmi_set_gpu_power_profile(args.gpu, 0, power_profile_mask)
reset_profile_results['power_profile'] = 'Successfully reset Power Profile to default (bootup default)'
except amdsmi_exception.AmdSmiLibraryException as e:
if e.get_error_code() == amdsmi_interface.amdsmi_wrapper.AMDSMI_STATUS_NO_PERM:
raise PermissionError('Command requires elevation') from e
reset_profile_results['power_profile'] = f"[{e.get_error_info(detailed=False)}] Unable to reset Power Profile to default (bootup default)"
logging.debug("Failed to reset power profile on gpu %s | %s", gpu_id, e.get_error_info())
self.logger.store_output(args.gpu, 'reset_profile', reset_profile_results)
self.logger.print_output()
self.logger.clear_multiple_devices_output()
return
if args.xgmierr:
try:
amdsmi_interface.amdsmi_reset_gpu_xgmi_error(args.gpu)
result = 'Successfully reset XGMI Error count'
except amdsmi_exception.AmdSmiLibraryException as e:
if e.get_error_code() == amdsmi_interface.amdsmi_wrapper.AMDSMI_STATUS_NO_PERM:
raise PermissionError('Command requires elevation') from e
logging.debug("Failed to reset xgmi error count on gpu %s | %s", gpu_id, e.get_error_info())
result = f"[{e.get_error_info(detailed=False)}] Unable to reset XGMI Error count"
self.logger.store_output(args.gpu, 'reset_xgmi_err', result)
self.logger.print_output()
self.logger.clear_multiple_devices_output()
return
self.logger.store_output(args.gpu, 'reset_xgmi_err', result)
self.logger.print_output()
self.logger.clear_multiple_devices_output()
return
if args.perf_determinism:
try:
level_auto = amdsmi_interface.AmdSmiDevPerfLevel.AUTO
amdsmi_interface.amdsmi_set_gpu_perf_level(args.gpu, level_auto)
result = 'Successfully reset Performance Level to default (auto)'
except amdsmi_exception.AmdSmiLibraryException as e:
if e.get_error_code() == amdsmi_interface.amdsmi_wrapper.AMDSMI_STATUS_NO_PERM:
raise PermissionError('Command requires elevation') from e
logging.debug("Failed to set perf level on gpu %s | %s", gpu_id, e.get_error_info())
result = f"[{e.get_error_info(detailed=False)}] Unable to reset Performance Level to default (auto)"
self.logger.store_output(args.gpu, 'reset_perf_determinism', result)
self.logger.print_output()
self.logger.clear_multiple_devices_output()
return
self.logger.store_output(args.gpu, 'reset_perf_determinism', result)
self.logger.print_output()
self.logger.clear_multiple_devices_output()
return
if args.power_cap:
final_output = {"ppt0": "[AMDSMI_STATUS_NOT_SUPPORTED] Unable to reset to default power cap", "ppt1": "[AMDSMI_STATUS_NOT_SUPPORTED] Unable to reset to default power cap"}
power_limit_types = {}
for power_type in amdsmi_interface.AmdSmiPowerCapType:
# Strip 'AMDSMI_POWER_CAP_TYPE_' prefix and convert to lowercase
key = power_type.name.replace('AMDSMI_POWER_CAP_TYPE_', '').lower()
power_limit_types[key] = "N/A"
current_sensor_num = 0
try:
power_cap_types = amdsmi_interface.amdsmi_get_supported_power_cap(args.gpu)
for sensor in power_cap_types['sensor_inds']:
current_sensor_num = sensor
power_cap_info = amdsmi_interface.amdsmi_get_power_cap_info(args.gpu, sensor)
logging.debug(f"Power cap info for gpu {gpu_id} ppt{sensor} | {power_cap_info}")
default_power_cap_in_mw = power_cap_info["default_power_cap"]
default_power_cap_in_w = self.helpers.convert_SI_unit(default_power_cap_in_mw, AMDSMIHelpers.SI_Unit.MICRO)
current_power_cap_in_mw = power_cap_info["power_cap"]
current_power_cap_in_w = self.helpers.convert_SI_unit(current_power_cap_in_mw, AMDSMIHelpers.SI_Unit.MICRO)
sensor_name = power_cap_types['sensor_types'][sensor]
# Strip 'AMDSMI_POWER_CAP_TYPE_' prefix and convert to lowercase
sensor_key = sensor_name.name.replace('AMDSMI_POWER_CAP_TYPE_', '').lower()
power_limit_types[sensor_key] = (default_power_cap_in_w, current_power_cap_in_w)
amdsmi_interface.amdsmi_set_power_cap(args.gpu, sensor, default_power_cap_in_mw)
final_output[f"ppt{current_sensor_num}"] = f"Successfully reset power cap to {default_power_cap_in_w}W"
except amdsmi_exception.AmdSmiLibraryException as e:
if e.get_error_code() == amdsmi_interface.amdsmi_wrapper.AMDSMI_STATUS_NO_PERM:
raise PermissionError('Command requires elevation') from e
final_output[f"ppt{current_sensor_num}"] = f"[{e.get_error_info(detailed=False)}] Unable to reset cap to default power cap"
self.logger.store_output(args.gpu, 'powercap', final_output)
if multiple_devices:
self.logger.store_multiple_device_output()
return
self.logger.print_output()
self.logger.clear_multiple_devices_output()
#######################
# BM commands - END #
#######################
if args.clean_local_data:
try:
amdsmi_interface.amdsmi_clean_gpu_local_data(args.gpu)
result = 'Successfully clean GPU local data'
except amdsmi_exception.AmdSmiLibraryException as e:
if e.get_error_code() == amdsmi_interface.amdsmi_wrapper.AMDSMI_STATUS_NO_PERM:
raise PermissionError('Command requires elevation') from e
result = f"[{e.get_error_info(detailed=False)}] Unable to clean local data"
self.logger.store_output(args.gpu, 'clean_local_data', result)
self.logger.print_output()
self.logger.clear_multiple_devices_output()
return
self.logger.store_output(args.gpu, 'clean_local_data', result)
self.logger.print_output()
self.logger.clear_multiple_devices_output()
return
# Adding to VMs since, they should also support same reload as baremetal
if args.reload_driver:
# Check permissions BEFORE starting any processes
# Required to avoid permission errors when starting the progress bar
try:
if os.geteuid() != 0:
result = "[AMDSMI_STATUS_NO_PERM] Command requires elevation"
self.logger.store_output(args.gpu, 'reload_driver', result)
self.logger.print_output()
self.logger.clear_multiple_devices_output()
raise PermissionError('Command requires elevation')
except AttributeError:
pass # os.geteuid() not available on Windows
lock = multiprocessing.Lock()
lock.acquire()
is_lock_released = False
progress_process = None
try:
self.helpers.increment_set_count()
set_count = self.helpers.get_set_count()
if set_count == 1:
self.helpers.confirm_gpu_driver_reload_warning()
# Start progress bar in separate process
string_out = f"Reloading driver for all AMD GPUs:"
progress_process = multiprocessing.Process(
target=self.helpers.showProgressbar,
args=(string_out, 140, True)
)
progress_process.start()
# Perform the actual driver reload (this is where permission error occurs)
amdsmi_interface.amdsmi_gpu_driver_reload()
# If we get here, operation was successful
self.helpers.assign_previous_set_success_check(amdsmi_interface.amdsmi_wrapper.AMDSMI_STATUS_SUCCESS)
result = "Successfully reloaded driver"
else:
if self.helpers.get_previous_set_success_check() == amdsmi_interface.amdsmi_wrapper.AMDSMI_STATUS_SUCCESS:
result = "Successfully reloaded driver"
elif self.helpers.get_previous_set_success_check() == amdsmi_interface.amdsmi_wrapper.AMDSMI_STATUS_NO_PERM:
result = "[AMDSMI_STATUS_NO_PERM] Command requires elevation"
raise PermissionError('Command requires elevation')
else:
previous_check = self.helpers.get_previous_set_success_check()
temp_exception = amdsmi_exception.AmdSmiLibraryException(previous_check)
str_out = temp_exception.get_error_info(detailed=False)
result = f"[{str_out}] Unable to successfully restart driver"
except amdsmi_exception.AmdSmiLibraryException as e:
# Handle permission error FIRST, before any cleanup
self.helpers.assign_previous_set_success_check(e.get_error_code())
if e.get_error_code() == amdsmi_interface.amdsmi_wrapper.AMDSMI_STATUS_NO_PERM:
self.helpers.assign_previous_set_success_check(amdsmi_interface.amdsmi_wrapper.AMDSMI_STATUS_NO_PERM)
result = f"[{e.get_error_info(detailed=False)}] Command requires elevation"
# Clean termination of progress bar
if progress_process and progress_process.is_alive():
progress_process.terminate()
progress_process.join(timeout=0.1) # Wait max 0.1 second
if progress_process.is_alive():
progress_process.kill() # Force kill if needed
print("\n") # Clean up progress bar line
# Store result and exit early
self.logger.store_output(args.gpu, 'reload_driver', result)
self.logger.print_output()
self.logger.clear_multiple_devices_output()
if not is_lock_released:
lock.release()
is_lock_released = True
raise PermissionError('Command requires elevation') from e
else:
# Handle other errors
self.helpers.assign_previous_set_success_check(e.get_error_code())
result = f"[{e.get_error_info(detailed=False)}] Unable to successfully restart driver"
finally:
# Always clean up progress bar process
if progress_process and progress_process.is_alive():
progress_process.terminate()
progress_process.join(timeout=0.1)
if progress_process.is_alive():
progress_process.kill()
print("\n") # Clean up progress bar line
# Always release lock
if not is_lock_released:
lock.release()
is_lock_released = True
# Store and print result
self.logger.store_output(args.gpu, 'reload_driver', result)
self.logger.print_output()
self.logger.clear_multiple_devices_output()
return
def monitor(self, args, multiple_devices=False, watching_output=False, gpu=None,
watch=None, watch_time=None, iterations=None, power_usage=None,
temperature=None, base_board_temps=None, gpu_board_temps=None,
gfx_util=None, mem_util=None, encoder=None, decoder=None,
ecc=None, vram_usage=None, pcie=None, process=None,
violation=None):
""" Populate a table with each GPU as an index to rows of targeted data
Args:
args (Namespace): Namespace containing the parsed CLI args
multiple_devices (bool, optional): True if checking for multiple devices. Defaults to False.
gpu (device_handle, optional): device_handle for target device. Defaults to None.
watch (bool, optional): Value override for args.watch. Defaults to None.
watch_time (int, optional): Value override for args.watch_time. Defaults to None.
iterations (int, optional): Value override for args.iterations. Defaults to None.
power_usage (bool, optional): Value override for args.power_usage. Defaults to None.
temperature (bool, optional): Value override for args.temperature. Defaults to None.
base_board_temps (bool, optional): Value override for args.base_board_temps. Defaults to None.
gpu_board_temps (bool, optional): Value override for args.gpu_board_temps. Defaults to None.
gfx (bool, optional): Value override for args.gfx. Defaults to None.
mem_util (bool, optional): Value override for args.mem. Defaults to None.
encoder (bool, optional): Value override for args.encoder. Defaults to None.
decoder (bool, optional): Value override for args.decoder. Defaults to None.
ecc (bool, optional): Value override for args.ecc. Defaults to None.
vram_usage (bool, optional): Value override for args.vram_usage. Defaults to None.
pcie (bool, optional): Value override for args.pcie. Defaults to None.
process (bool, optional): Value override for args.process. Defaults to None.
violation (bool, optional): Value override for args.violation. Defaults to None.
Raises:
ValueError: Value error if no gpu value is provided
IndexError: Index error if gpu list is empty
Return:
Nothing
"""
# Set args.* to passed in arguments
if gpu:
args.gpu = gpu
if watch:
args.watch = watch
if watch_time:
args.watch_time = watch_time
if iterations:
args.iterations = iterations
# monitor args
if power_usage:
args.power_usage = power_usage
if temperature:
args.temperature = temperature
if base_board_temps:
args.base_board_temps = base_board_temps
if gpu_board_temps:
args.gpu_board_temps = gpu_board_temps
if gfx_util:
args.gfx = gfx_util
if mem_util:
args.mem = mem_util
if encoder:
args.encoder = encoder
if decoder:
args.decoder = decoder
if ecc:
args.ecc = ecc
if vram_usage:
args.vram_usage = vram_usage
if pcie:
args.pcie = pcie
if process:
args.process = process
if not self.helpers.is_virtual_os():
if violation:
args.violation = violation
else:
args.violation = False # Disable violation for virtual OS
# Handle No GPU passed
if args.gpu == None:
args.gpu = self.device_handles
if not self.group_check_printed:
self.helpers.check_required_groups(check_render=True, check_video=False)
self.group_check_printed = True
# If all arguments are False, the print all values
# Don't include process in this logic as it's an optional edge case
if not any([args.power_usage, args.temperature, args.base_board_temps,
args.gpu_board_temps, args.gfx, args.mem, args.encoder,
args.decoder, args.ecc, args.vram_usage, args.pcie,
args.violation]):
args.power_usage = args.temperature = args.gfx = args.mem = \
args.encoder = args.decoder = args.vram_usage = True
# set extra args for default output filtering
args.default_output = True
else:
if not hasattr(args, 'default_output'):
args.default_output = False
# Handle watch logic, will only enter this block once
if args.watch:
self.helpers.handle_watch(args=args, subcommand=self.monitor, logger=self.logger)
return
# Handle multiple GPUs
if isinstance(args.gpu, list):
if len(args.gpu) > 1:
# Deepcopy gpus as recursion will destroy the gpu list
stored_gpus = []
for gpu in args.gpu:
stored_gpus.append(gpu)
# Store output from multiple devices without printing to console
for device_handle in args.gpu:
self.monitor(args, multiple_devices=True, watching_output=watching_output, gpu=device_handle)
# Reload original gpus
args.gpu = stored_gpus
dual_csv_output = False
if args.process:
if self.logger.is_csv_format():
dual_csv_output = True
# Flush the output
self.logger.print_output(multiple_device_enabled=True,
watching_output=watching_output,
tabular=True,
dual_csv_output=dual_csv_output)
# Add output to total watch output and clear multiple device output
if watching_output:
self.logger.store_watch_output(multiple_device_enabled=True)
return
elif len(args.gpu) == 1:
args.gpu = args.gpu[0]
else:
raise IndexError("args.gpu should not be an empty list")
monitor_values = {}
# Get gpu_id for logging
gpu_id = self.helpers.get_gpu_id_from_device_handle(args.gpu)
# Reset the table header and store the timestamp if watch output is enabled
self.logger.table_header = 'GPU'
if watching_output:
self.logger.store_output(args.gpu, 'timestamp', int(time.time()))
self.logger.table_header = 'TIMESTAMP'.rjust(10) + ' ' + self.logger.table_header
if args.loglevel == "DEBUG":
try:
# Get GPU Metrics table version
gpu_metric_version_info = amdsmi_interface.amdsmi_get_gpu_metrics_header_info(args.gpu)
gpu_metric_version_str = json.dumps(gpu_metric_version_info, indent=4)
logging.debug("GPU Metrics table Version for GPU %s | %s", gpu_id, gpu_metric_version_str)
except amdsmi_exception.AmdSmiLibraryException as e:
logging.debug("#4 - Unable to load GPU Metrics table version for %s | %s", gpu_id, e.get_error_info())
try:
# Get GPU Metrics table
gpu_metric_debug_info = amdsmi_interface.amdsmi_get_gpu_metrics_info(args.gpu)
except amdsmi_exception.AmdSmiLibraryException as e:
logging.debug("#5 - Unable to load GPU Metrics table for %s | %s", gpu_id, e.get_error_info())
is_partition_metrics = False # True if we get the metrics from xcp_metrics file (amdsmi_get_gpu_partition_metrics_info)
#get metric info only once per gpu, this will speed up data output
try:
# Get GPU Metrics table
gpu_metrics_info = amdsmi_interface.amdsmi_get_gpu_metrics_info(args.gpu)
if args.loglevel == "DEBUG":
gpu_metric_debug_info = json.dumps(gpu_metrics_info, indent=4)
logging.debug("GPU Metrics table for GPU %s | %s", gpu_id, gpu_metric_debug_info)
except amdsmi_exception.AmdSmiLibraryException as e:
gpu_metrics_info = amdsmi_interface._NA_amdsmi_get_gpu_metrics_info()
logging.debug("Unable to load GPU Metrics table for %s | %s", gpu_id, e.get_error_info())
# Workaround for XCP (partition) metrics not providing num_partition in v1.9+/v1.1+
# Provides original formatting for earlier metric versions
partition_metric_info = self.helpers._get_metric_version_and_partition_info(gpu_metrics_info, is_partition_metrics, gpu_id, args.gpu)
partition_id = partition_metric_info['partition_id']
num_partition = partition_metric_info['num_partition']
# Update logger for XCP display (only if applicable)
self.logger.table_header += 'XCP'.rjust(5, ' ')
self.logger.store_output(args.gpu, 'xcp', partition_id) # Store partition_id initially; can be updated via num_xcp
# Store the pcie_bw values due to possible increase in bandwidth due to repeated gpu_metrics calls
if args.pcie:
try:
pcie_info = amdsmi_interface.amdsmi_get_pcie_info(args.gpu)['pcie_metric']
except amdsmi_exception.AmdSmiLibraryException as e:
pcie_info = "N/A"
logging.debug("Failed to get pci bandwidth on gpu %s | %s", gpu_id, e.get_error_info())
power_unit = 'W'
# Resume regular ordering of values
if args.power_usage:
try:
if gpu_metrics_info['current_socket_power'] != "N/A":
monitor_values['power_usage'] = gpu_metrics_info['current_socket_power']
else: # Fallback to average_socket_power for older gpu_metrics versions
monitor_values['power_usage'] = gpu_metrics_info['average_socket_power']
if self.logger.is_human_readable_format() and monitor_values['power_usage'] != "N/A":
monitor_values['power_usage'] = f"{monitor_values['power_usage']} {power_unit}"
if self.logger.is_json_format() and monitor_values['power_usage'] != "N/A":
monitor_values['power_usage'] = {"value" : monitor_values['power_usage'],
"unit" : power_unit}
except (KeyError, amdsmi_exception.AmdSmiLibraryException) as e:
monitor_values['power_usage'] = "N/A"
logging.debug("Failed to get power usage on gpu %s | %s", gpu_id, e)
self.logger.table_header += 'POWER'.rjust(7)
if args.power_usage and not args.default_output:
# Get Current Power Cap
try:
# assume that we're always asking for ppt0 for quick checks like this
power_cap_info = amdsmi_interface.amdsmi_get_power_cap_info(args.gpu, 0)
monitor_values['max_power'] = power_cap_info['power_cap'] # Get current power cap (`power_cap`) socket is set to
# `max_power_cap`, is the maximum value it can be set to
monitor_values['max_power'] = self.helpers.convert_SI_unit(monitor_values['max_power'], AMDSMIHelpers.SI_Unit.MICRO)
if self.logger.is_human_readable_format() and monitor_values['max_power'] != "N/A":
monitor_values['max_power'] = f"{monitor_values['max_power']} {power_unit}"
if self.logger.is_json_format() and monitor_values['max_power'] != "N/A":
monitor_values['max_power'] = {"value" : monitor_values['max_power'],
"unit" : power_unit}
except amdsmi_exception.AmdSmiLibraryException as e:
monitor_values['max_power'] = "N/A"
logging.debug("Failed to get power cap info for gpu %s | %s", gpu_id, e.get_error_info())
self.logger.table_header += 'PWR_CAP'.rjust(9)
if args.temperature:
try:
temperature = gpu_metrics_info['temperature_hotspot']
monitor_values['hotspot_temperature'] = temperature
except (KeyError, amdsmi_exception.AmdSmiLibraryException) as e:
monitor_values['hotspot_temperature'] = "N/A"
logging.debug("Failed to get hotspot temperature on gpu %s | %s", gpu_id, e)
try:
temperature = gpu_metrics_info['temperature_mem']
monitor_values['memory_temperature'] = temperature
except (KeyError, amdsmi_exception.AmdSmiLibraryException) as e:
monitor_values['memory_temperature'] = "N/A"
logging.debug("Failed to get memory temperature on gpu %s | %s", gpu_id, e)
temp_unit_human_readable = '\N{DEGREE SIGN}C'
temp_unit_json = 'C'
if monitor_values['hotspot_temperature'] != "N/A":
if self.logger.is_human_readable_format():
monitor_values['hotspot_temperature'] = f"{monitor_values['hotspot_temperature']} {temp_unit_human_readable}"
if self.logger.is_json_format():
monitor_values['hotspot_temperature'] = {"value" : monitor_values['hotspot_temperature'],
"unit" : temp_unit_json}
if monitor_values['memory_temperature'] != "N/A":
if self.logger.is_human_readable_format():
monitor_values['memory_temperature'] = f"{monitor_values['memory_temperature']} {temp_unit_human_readable}"
if self.logger.is_json_format():
monitor_values['memory_temperature'] = {"value" : monitor_values['memory_temperature'],
"unit" : temp_unit_json}
self.logger.table_header += 'GPU_T'.rjust(8)
self.logger.table_header += 'MEM_T'.rjust(8)
if args.gpu_board_temps:
try:
gpu_board_temp_dict = self.helpers.get_gpu_board_temperatures(args.gpu, gpu_id, self.logger)
temp_unit_json = 'C'
# Add GPU board sensor headers
if gpu_board_temp_dict:
for temp_sensor in sorted(gpu_board_temp_dict.keys()):
self.logger.table_header += f"{temp_sensor}".rjust(max(len(temp_sensor)+2, 7))
for temp_type, temp_value in gpu_board_temp_dict.items():
if self.logger.is_json_format() and isinstance(temp_value, dict):
temp_value['unit'] = temp_unit_json
monitor_values[temp_type] = temp_value
except Exception as e:
logging.debug("Failed to get GPU board temperatures on gpu %s | %s", gpu_id, e)
if args.base_board_temps:
try:
base_board_temp_dict = self.helpers.get_base_board_temperatures(args.gpu, gpu_id, self.logger)
temp_unit_json = 'C'
# Add base board sensor headers
if base_board_temp_dict:
for temp_sensor in sorted(base_board_temp_dict.keys()):
self.logger.table_header += f"{temp_sensor}".rjust(max(len(temp_sensor)+2, 7))
for temp_type, temp_value in base_board_temp_dict.items():
if self.logger.is_json_format() and isinstance(temp_value, dict):
temp_value['unit'] = temp_unit_json
monitor_values[temp_type] = temp_value
except Exception as e:
logging.debug("Failed to get base board temperatures on gpu %s | %s", gpu_id, e)
if args.gfx:
try:
gfx_clk = gpu_metrics_info['current_gfxclk']
monitor_values['gfx_clk'] = gfx_clk
freq_unit = 'MHz'
if gfx_clk != "N/A":
if self.logger.is_human_readable_format():
monitor_values['gfx_clk'] = f"{monitor_values['gfx_clk']} {freq_unit}"
if self.logger.is_json_format():
monitor_values['gfx_clk'] = {"value" : monitor_values['gfx_clk'],
"unit" : freq_unit}
except (KeyError, amdsmi_exception.AmdSmiLibraryException) as e:
monitor_values['gfx_clk'] = "N/A"
logging.debug("Failed to get gfx clock on gpu %s | %s", gpu_id, e)
self.logger.table_header += 'GFX_CLK'.rjust(10)
try:
gfx_util = gpu_metrics_info['average_gfx_activity']
activity_unit = '%'
if gfx_util != "N/A":
monitor_values['gfx'] = gfx_util
if self.logger.is_human_readable_format():
monitor_values['gfx'] = f"{monitor_values['gfx']} {activity_unit}"
if self.logger.is_json_format():
monitor_values['gfx'] = {"value" : monitor_values['gfx'],
"unit" : activity_unit}
except (KeyError, amdsmi_exception.AmdSmiLibraryException) as e:
monitor_values['gfx'] = "N/A"
logging.debug("Failed to get gfx utilization on gpu %s | %s", gpu_id, e)
self.logger.table_header += 'GFX%'.rjust(7)
if args.mem:
try:
mem_util = gpu_metrics_info['average_umc_activity']
activity_unit = '%'
if mem_util != "N/A":
monitor_values['mem'] = mem_util
if self.logger.is_human_readable_format():
monitor_values['mem'] = f"{monitor_values['mem']} {activity_unit}"
if self.logger.is_json_format():
monitor_values['mem'] = {"value" : monitor_values['mem'],
"unit" : activity_unit}
except (KeyError, amdsmi_exception.AmdSmiLibraryException) as e:
monitor_values['mem'] = "N/A"
logging.debug("Failed to get mem utilization on gpu %s | %s", gpu_id, e)
self.logger.table_header += 'MEM%'.rjust(7)
# don't populate mem clock on default output
if not args.default_output:
try:
mem_clock = gpu_metrics_info['current_uclk']
monitor_values['mem_clock'] = mem_clock
freq_unit = 'MHz'
if mem_clock != "N/A":
if self.logger.is_human_readable_format():
monitor_values['mem_clock'] = f"{monitor_values['mem_clock']} {freq_unit}"
if self.logger.is_json_format():
monitor_values['mem_clock'] = {"value" : monitor_values['mem_clock'],
"unit" : freq_unit}
except (KeyError, amdsmi_exception.AmdSmiLibraryException) as e:
monitor_values['mem_clock'] = "N/A"
logging.debug("Failed to get mem clock on gpu %s | %s", gpu_id, e)
self.logger.table_header += 'MEM_CLOCK'.rjust(11)
if args.encoder:
# TODO: The encoding utilization is in progress for Navi. Note: MI3x ASICs only support decoding.
try:
# Get List of vcn activity values
encoder_util = "N/A" # Not yet implemented
encoding_activity_avg = []
for value in encoder_util:
if isinstance(value, int):
encoding_activity_avg.append(value)
# Averaging the possible encoding activity values
if encoding_activity_avg:
encoding_activity_avg = round(sum(encoding_activity_avg) / len(encoding_activity_avg))
else:
encoding_activity_avg = "N/A"
monitor_values['encoder'] = encoding_activity_avg
activity_unit = '%'
if monitor_values['encoder'] != "N/A":
if self.logger.is_human_readable_format():
monitor_values['encoder'] = f"{monitor_values['encoder']} {activity_unit}"
if self.logger.is_json_format():
monitor_values['encoder'] = {"value" : monitor_values['encoder'],
"unit" : activity_unit}
except amdsmi_exception.AmdSmiLibraryException as e:
monitor_values['encoder'] = "N/A"
logging.debug("Failed to get encoder utilization on gpu %s | %s", gpu_id, e.get_error_info())
self.logger.table_header += 'ENC%'.rjust(7)
if args.decoder:
try:
# Get List of vcn activity values
# Note: MI3x ASICs only support decoding, so the vcn_activity/vcn_busy
# is used for decoding activity.
decoder_util = gpu_metrics_info['vcn_activity']
if (gpu_metrics_info['vcn_activity'][0] == "N/A" and
gpu_metrics_info['xcp_stats.vcn_busy'][partition_id][0] != "N/A"):
decoder_util = gpu_metrics_info['xcp_stats.vcn_busy'][partition_id]
decoding_activity_avg = self.helpers.average_flattened_ints(decoder_util, context="decoder_util")
monitor_values['decoder'] = decoding_activity_avg
activity_unit = '%'
if monitor_values['decoder'] != "N/A":
if self.logger.is_human_readable_format():
monitor_values['decoder'] = f"{monitor_values['decoder']} {activity_unit}"
if self.logger.is_json_format():
monitor_values['decoder'] = {"value" : monitor_values['decoder'],
"unit" : activity_unit}
except (KeyError, amdsmi_exception.AmdSmiLibraryException) as e:
monitor_values['decoder'] = "N/A"
logging.debug("Failed to get decoder utilization on gpu %s | %s", gpu_id, e)
self.logger.table_header += 'DEC%'.rjust(7)
if (args.encoder or args.decoder) and not args.default_output:
try:
vclock = gpu_metrics_info['current_vclk0']
monitor_values['vclock'] = vclock
freq_unit = 'MHz'
if vclock != "N/A":
if self.logger.is_human_readable_format():
monitor_values['vclock'] = f"{monitor_values['vclock']} {freq_unit}"
if self.logger.is_json_format():
monitor_values['vclock'] = {"value" : monitor_values['vclock'],
"unit" : freq_unit}
except (KeyError, amdsmi_exception.AmdSmiLibraryException) as e:
monitor_values['vclock'] = "N/A"
logging.debug("Failed to get dclock on gpu %s | %s", gpu_id, e)
self.logger.table_header += 'VCLOCK'.rjust(10)
try:
dclock = gpu_metrics_info['current_dclk0']
monitor_values['dclock'] = dclock
freq_unit = 'MHz'
if dclock != "N/A":
if self.logger.is_human_readable_format():
monitor_values['dclock'] = f"{monitor_values['dclock']} {freq_unit}"
if self.logger.is_json_format():
monitor_values['dclock'] = {"value" : monitor_values['dclock'],
"unit" : freq_unit}
except (KeyError, amdsmi_exception.AmdSmiLibraryException) as e:
monitor_values['dclock'] = "N/A"
logging.debug("Failed to get dclock on gpu %s | %s", gpu_id, e)
self.logger.table_header += 'DCLOCK'.rjust(10)
if args.ecc:
try:
ecc = amdsmi_interface.amdsmi_get_gpu_total_ecc_count(args.gpu)
monitor_values['single_bit_ecc'] = ecc['correctable_count']
monitor_values['double_bit_ecc'] = ecc['uncorrectable_count']
except amdsmi_exception.AmdSmiLibraryException as e:
monitor_values['ecc'] = "N/A"
logging.debug("Failed to get ecc on gpu %s | %s", gpu_id, e.get_error_info())
self.logger.table_header += 'SINGLE_ECC'.rjust(12)
self.logger.table_header += 'DOUBLE_ECC'.rjust(12)
try:
pcie_metric = amdsmi_interface.amdsmi_get_pcie_info(args.gpu)['pcie_metric']
logging.debug("PCIE Metric for %s | %s", gpu_id, pcie_metric)
monitor_values['pcie_replay'] = pcie_metric['pcie_replay_count']
except amdsmi_exception.AmdSmiLibraryException as e:
monitor_values['pcie_replay'] = "N/A"
logging.debug("Failed to get gpu_metrics pcie replay counter on gpu %s | %s", gpu_id, e.get_error_info())
if monitor_values['pcie_replay'] == "N/A":
try:
pcie_replay = amdsmi_interface.amdsmi_get_gpu_pci_replay_counter(args.gpu)
monitor_values['pcie_replay'] = pcie_replay
except amdsmi_exception.AmdSmiLibraryException as e:
logging.debug("Failed to get sysfs pcie replay counter on gpu %s | %s", gpu_id, e.get_error_info())
self.logger.table_header += 'PCIE_REPLAY'.rjust(13)
if args.vram_usage and not args.default_output:
mem_type, mem_type_name = self.helpers.get_apu_memory_type_and_name(args.gpu, gpu_id)
try:
mem_used = amdsmi_interface.amdsmi_get_gpu_memory_usage(args.gpu, mem_type) // (1024*1024)
mem_total = amdsmi_interface.amdsmi_get_gpu_memory_total(args.gpu, mem_type) // (1024*1024)
monitor_values['vram_used'] = mem_used
monitor_values['vram_free'] = mem_total - mem_used
monitor_values['vram_total'] = mem_total
if mem_total != 0:
monitor_values['vram_percent'] = round ((mem_used / mem_total) * 100, 2)
else:
monitor_values['vram_percent'] = "N/A"
mem_usage_unit = "MB"
mem_percent_unit = "%"
if self.logger.is_human_readable_format():
monitor_values['vram_used'] = f"{monitor_values['vram_used']} {mem_usage_unit}"
monitor_values['vram_free'] = f"{monitor_values['vram_free']} {mem_usage_unit}"
monitor_values['vram_total'] = f"{monitor_values['vram_total']} {mem_usage_unit}"
monitor_values['vram_percent'] = f"{monitor_values['vram_percent']} {mem_percent_unit}"
if self.logger.is_json_format():
monitor_values['vram_used'] = {"value" : monitor_values['vram_used'],
"unit" : mem_usage_unit}
monitor_values['vram_free'] = {"value" : monitor_values['vram_free'],
"unit" : mem_usage_unit}
monitor_values['vram_total'] = {"value" : monitor_values['vram_total'],
"unit" : mem_usage_unit}
monitor_values['vram_percent'] = {"value" : monitor_values['vram_percent'],
"unit" : mem_percent_unit}
except amdsmi_exception.AmdSmiLibraryException as e:
monitor_values['vram_used'] = "N/A"
monitor_values['vram_free'] = "N/A"
monitor_values['vram_total'] = "N/A"
monitor_values['vram_percent'] = "N/A"
logging.debug("Failed to get %s memory usage on gpu %s | %s", mem_type_name.lower(), gpu_id, e.get_error_info())
# Use appropriate headers based on memory type
self.logger.table_header += f'{mem_type_name}_USED'.rjust(11)
self.logger.table_header += f'{mem_type_name}_FREE'.rjust(12)
self.logger.table_header += f'{mem_type_name}_TOTAL'.rjust(12)
self.logger.table_header += f'{mem_type_name}%'.rjust(9)
if args.vram_usage and args.default_output:
mem_type, mem_type_name = self.helpers.get_apu_memory_type_and_name(args.gpu, gpu_id)
try:
mem_used = amdsmi_interface.amdsmi_get_gpu_memory_usage(args.gpu, mem_type) // (1024*1024)
mem_total = amdsmi_interface.amdsmi_get_gpu_memory_total(args.gpu, mem_type) // (1024*1024)
mem_usage_unit = "GB"
if self.logger.is_json_format():
monitor_values['vram_used'] = {"value" : round(mem_used/1024,1),
"unit" : mem_usage_unit}
monitor_values['vram_total'] = {"value" : round(mem_total/1024,1),
"unit" : mem_usage_unit}
elif self.logger.is_csv_format():
monitor_values['vram_used'] = round(mem_used/1024,1)
monitor_values['vram_total'] = round(mem_total/1024,1)
else:
monitor_values['vram_usage'] = f"{mem_used/1024:5.1f}/{mem_total/1024:5.1f} {mem_usage_unit}".rjust(16,' ')
except amdsmi_exception.AmdSmiLibraryException as e:
if self.logger.is_json_format():
monitor_values['vram_used'] = "N/A"
monitor_values['vram_total'] = "N/A"
else:
monitor_values['vram_usage'] = "N/A"
logging.debug("Failed to get %s memory usage on gpu %s | %s", mem_type_name.lower(), gpu_id, e.get_error_info())
# Use appropriate header based on memory type
header_name = f'{mem_type_name}_USAGE'
self.logger.table_header += header_name.rjust(16)
if args.pcie:
if pcie_info != "N/A":
pcie_bw_unit = 'Mb/s'
monitor_values['pcie_bw'] = self.helpers.unit_format(self.logger, pcie_info['pcie_bandwidth'], pcie_bw_unit)
else:
monitor_values['pcie_bw'] = pcie_info
self.logger.table_header += 'PCIE_BW'.rjust(12)
# initialize dual_csv_format; applicable to process only
dual_csv_output = False
# Store process list separately
if args.process:
# Populate initial processes
try:
process_list = amdsmi_interface.amdsmi_get_gpu_process_list(args.gpu)
except amdsmi_exception.AmdSmiLibraryException as e:
if e.get_error_code() == amdsmi_interface.amdsmi_wrapper.AMDSMI_STATUS_NO_PERM:
raise PermissionError('Command requires elevation') from e
logging.debug("Failed to get process list for gpu %s | %s", gpu_id, e.get_error_info())
raise e
try:
num_compute_units = amdsmi_interface.amdsmi_get_gpu_asic_info(args.gpu)['num_compute_units']
except (KeyError, amdsmi_exception.AmdSmiLibraryException) as e:
num_compute_units = "N/A"
logging.debug("Failed to get num compute units for gpu %s | %s", gpu_id, e.get_error_info())
# Clean processes dictionary
filtered_process_values = []
for process_info in process_list:
process_info.pop('engine_usage') # Remove 'engine_usage' value
process_info['mem_usage'] = process_info.pop('mem')
process_info['cu_occupancy'] = process_info.pop('cu_occupancy')
process_info['evicted_time'] = process_info.pop('evicted_time')
memory_usage_unit = "B"
evicted_time_unit = "ms"
if self.logger.is_human_readable_format():
process_info['mem_usage'] = self.helpers.convert_bytes_to_readable(process_info['mem_usage'])
for usage_metric in process_info['memory_usage']:
process_info["memory_usage"][usage_metric] = self.helpers.convert_bytes_to_readable(process_info["memory_usage"][usage_metric])
memory_usage_unit = ""
process_info['mem_usage'] = self.helpers.unit_format(self.logger,
process_info['mem_usage'],
memory_usage_unit)
process_info['evicted_time'] = self.helpers.unit_format(self.logger,
process_info['evicted_time'],
evicted_time_unit)
for usage_metric in process_info['memory_usage']:
process_info['memory_usage'][usage_metric] = self.helpers.unit_format(self.logger,
process_info['memory_usage'][usage_metric],
memory_usage_unit)
if 'cu_occupancy' in process_info:
try:
cu_occupancy = process_info['cu_occupancy']
if num_compute_units != "N/A" and num_compute_units > 0 and cu_occupancy != "N/A":
cu_percentage = round((cu_occupancy / num_compute_units) * 100, 1)
process_info['cu_occupancy'] = self.helpers.unit_format(self.logger,
cu_percentage,
'%')
else:
process_info['cu_occupancy'] = "N/A"
except Exception as e:
process_info['cu_occupancy'] = "N/A"
logging.debug("Failed to calculate cu_occupancy percentage for GPU %s | %s", gpu_id, str(e))
filtered_process_values.append({'process_info': process_info})
# If no processes are populated then we populate an N/A placeholder
if not filtered_process_values:
logging.debug("Monitor - Failed to detect any process on gpu %s", gpu_id)
filtered_process_values.append({'process_info': "N/A"})
for index, process in enumerate(filtered_process_values):
if process['process_info'] == "N/A":
filtered_process_values[index]['process_info'] = "No running processes detected"
# Build the process table's title and header
self.logger.secondary_table_title = "PROCESS INFO"
self.logger.secondary_table_header = 'GPU'.rjust(3) + "NAME".rjust(19) + "PID".rjust(9) + "GTT_MEM".rjust(10) + \
"CPU_MEM".rjust(10) + "VRAM_MEM".rjust(10) + "MEM_USG".rjust(10) + "CU%".rjust(9) + "EVICT".rjust(10)
if watching_output:
self.logger.secondary_table_header = 'TIMESTAMP'.rjust(10) + ' ' + self.logger.secondary_table_header
logging.debug(f"Monitor - Process Info for GPU {gpu_id} | {filtered_process_values}")
if self.logger.is_json_format():
self.logger.store_output(args.gpu, 'process_list', filtered_process_values)
if self.logger.is_human_readable_format():
# Print out process in flattened format
# The logger detects if process list is present and pulls it out and prints
# that table with timestamp, gpu, and prints headers separately
self.logger.store_output(args.gpu, 'process_list', filtered_process_values)
if self.logger.is_csv_format():
dual_csv_output = True
# The logger detects if process list is present and pulls it out and prints
# that table with timestamp, gpu, and prints headers separately
self.logger.store_output(args.gpu, 'process_list', filtered_process_values)
###################
### XCP Metrics ###
###################
# Must come after process list - XCP detail is a multi-dimensional array, which is displayed
# in tabular format with XCP values for same gpu shown on muliple lines.
if args.violation:
violation_status = {
"pviol": "N/A",
"tviol": "N/A",
"tviol_active": "N/A",
"phot_tviol": "N/A",
"vr_tviol": "N/A",
"hbm_tviol": "N/A",
"gfx_clkviol": "N/A",
"gfxclk_pviol": "N/A",
"gfxclk_tviol": "N/A",
"gfxclk_totalviol": "N/A",
"low_utilviol": "N/A"
}
try:
violations = amdsmi_interface.amdsmi_get_violation_status(args.gpu)
violation_status['pviol'] = violations['per_ppt_pwr']
violation_status['tviol'] = violations['per_socket_thrm']
violation_status['tviol_active'] = violations['active_socket_thrm']
violation_status['phot_tviol'] = violations['per_prochot_thrm']
violation_status['vr_tviol'] = violations['per_vr_thrm']
violation_status['hbm_tviol'] = violations['per_hbm_thrm']
violation_status['gfx_clkviol'] = violations['per_gfx_clk_below_host_limit']
violation_status['gfxclk_pviol'] = violations['per_gfx_clk_below_host_limit_pwr']
violation_status['gfxclk_tviol'] = violations['per_gfx_clk_below_host_limit_thm']
violation_status['gfxclk_totalviol'] = violations['per_gfx_clk_below_host_limit_total']
violation_status['low_utilviol'] = violations['per_low_utilization']
except amdsmi_exception.AmdSmiLibraryException as e:
monitor_values['pviol'] = violation_status['pviol']
monitor_values['tviol'] = violation_status['tviol']
monitor_values['tviol_active'] = violation_status['tviol_active']
monitor_values['phot_tviol'] = violation_status['phot_tviol']
monitor_values['vr_tviol'] = violation_status['vr_tviol']
monitor_values['hbm_tviol'] = violation_status['hbm_tviol']
monitor_values['gfx_clkviol'] = violation_status['gfx_clkviol']
monitor_values['gfxclk_pviol'] = violation_status['gfxclk_pviol']
monitor_values['gfxclk_tviol'] = violation_status['gfxclk_tviol']
monitor_values['gfxclk_totalviol'] = violation_status['gfxclk_totalviol']
monitor_values['low_utilviol'] = violation_status['low_utilviol']
logging.debug("Failed to get violation status on gpu %s | %s", gpu_id, e.get_error_info())
violation_status_unit = "%"
kPVIOL_MAX_WIDTH = 7
kTVIOL_MAX_WIDTH = 7
kTVIOL_ACTIVE_MAX_WIDTH = 14
kPHOT_MAX_WIDTH = 12
kVR_MAX_WIDTH = 10
kHBM_MAX_WIDTH = 11
kGFXC_MAX_WIDTH = 13
kGFXC_PVIOL_MAX_WIDTH = 58
kGFXC_TVIOL_MAX_WIDTH = kGFXC_PVIOL_MAX_WIDTH
kGFXC_TOTALVIOL_MAX_WIDTH = kGFXC_PVIOL_MAX_WIDTH
kLOW_UTILVIOL_MAX_WIDTH = kGFXC_PVIOL_MAX_WIDTH
for key, value in violation_status.items():
if not isinstance(value, list):
if value != "N/A":
if key == 'tviol_active' or key == 'xcp':
monitor_values[key] = value
else:
monitor_values[key] = self.helpers.unit_format(self.logger, violation_status[key], violation_status_unit)
else:
monitor_values[key] = violation_status[key]
else:
if num_partition != "N/A":
# these are one after another, in order to display each in sub-sections
new_xcp_dict = {}
for current_xcp in range(num_partition):
new_xcp_dict[f"xcp_{current_xcp}"] = self.helpers.unit_format(self.logger, value[current_xcp], "%")
monitor_values[key] = new_xcp_dict
else:
monitor_values[key] = value[0] if value else "N/A"
# save deep copy of monitor values, used later to grab xcp specific values
monitor_values_deepcopy = copy.deepcopy(monitor_values)
self.logger.table_header += 'PVIOL'.rjust(kPVIOL_MAX_WIDTH, ' ')
self.logger.table_header += 'TVIOL'.rjust(kTVIOL_MAX_WIDTH, ' ')
self.logger.table_header += 'TVIOL_ACTIVE'.rjust(kTVIOL_ACTIVE_MAX_WIDTH, ' ')
self.logger.table_header += 'PHOT_TVIOL'.rjust(kPHOT_MAX_WIDTH, ' ')
self.logger.table_header += 'VR_TVIOL'.rjust(kVR_MAX_WIDTH, ' ')
self.logger.table_header += 'HBM_TVIOL'.rjust(kHBM_MAX_WIDTH, ' ')
self.logger.table_header += 'GFX_CLKVIOL'.rjust(kGFXC_MAX_WIDTH, ' ')
self.logger.table_header += 'GFXCLK_PVIOL'.rjust(kGFXC_PVIOL_MAX_WIDTH, ' ')
self.logger.table_header += 'GFXCLK_TVIOL'.rjust(kGFXC_TVIOL_MAX_WIDTH, ' ')
self.logger.table_header += 'GFXCLK_TOTALVIOL'.rjust(kGFXC_TOTALVIOL_MAX_WIDTH, ' ')
self.logger.table_header += 'LOW_UTILVIOL'.rjust(kLOW_UTILVIOL_MAX_WIDTH, ' ')
# Print/capture by XCPs
if num_partition != "N/A" and partition_id == 0:
current_xcp = 0
while (current_xcp in range(num_partition) or current_xcp == 0):
if not multiple_devices and watching_output and current_xcp == 0:
# Need to clear output for single device, otherwise while watching output
# XCP detail will continue stacking on top of each other
self.logger.clear_multiple_devices_output()
if watching_output:
self.logger.store_output(args.gpu, 'timestamp', int(time.time()))
if current_xcp != 0: # set all other values without XCP stats to N/A
self.logger.store_output(args.gpu, 'xcp', current_xcp)
monitor_values['pviol'] = "N/A"
monitor_values['tviol'] = "N/A"
monitor_values['tviol_active'] = "N/A"
monitor_values['phot_tviol'] = "N/A"
monitor_values['vr_tviol'] = "N/A"
monitor_values['hbm_tviol'] = "N/A"
monitor_values['gfx_clkviol'] = "N/A"
for k, _ in monitor_values.items(): # change other keys to "N/A" since we should have all applicable XCP stats
# eg. amd-smi monitor -p -t -V should only show XCP info for violations
# below primary device
if k != 'xcp' and k not in ['gfxclk_pviol', 'gfxclk_tviol', 'gfxclk_totalviol', 'low_utilviol']:
monitor_values[k] = "N/A"
if isinstance(monitor_values_deepcopy['gfxclk_pviol'], dict):
monitor_values['gfxclk_pviol'] = monitor_values_deepcopy['gfxclk_pviol'][f"xcp_{current_xcp}"]
if isinstance(monitor_values_deepcopy['gfxclk_tviol'], dict):
monitor_values['gfxclk_tviol'] = monitor_values_deepcopy['gfxclk_tviol'][f"xcp_{current_xcp}"]
if isinstance(monitor_values_deepcopy['gfxclk_totalviol'], dict):
monitor_values['gfxclk_totalviol'] = monitor_values_deepcopy['gfxclk_totalviol'][f"xcp_{current_xcp}"]
if isinstance(monitor_values_deepcopy['low_utilviol'], dict):
monitor_values['low_utilviol'] = monitor_values_deepcopy['low_utilviol'][f"xcp_{current_xcp}"]
if self.logger.is_human_readable_format():
monitor_values['pviol'] = monitor_values['pviol']
monitor_values['tviol'] = monitor_values['tviol']
monitor_values['phot_tviol'] = monitor_values['phot_tviol']
monitor_values['vr_tviol'] = monitor_values['vr_tviol']
monitor_values['hbm_tviol'] = monitor_values['hbm_tviol']
monitor_values['gfx_clkviol'] = monitor_values['gfx_clkviol']
monitor_values['gfxclk_pviol'] = str(monitor_values['gfxclk_pviol']).replace('\'', '')
monitor_values['gfxclk_tviol'] = str(monitor_values['gfxclk_tviol']).replace('\'', '')
monitor_values['gfxclk_totalviol'] = str(monitor_values['gfxclk_totalviol']).replace('\'', '')
monitor_values['low_utilviol'] = str(monitor_values['low_utilviol']).replace('\'', '')
self.logger.store_output(args.gpu, 'values', monitor_values)
self.logger.store_multiple_device_output()
current_xcp += 1
else:
self.logger.store_output(args.gpu, 'xcp', partition_id)
self.logger.store_output(args.gpu, 'values', monitor_values)
# Store typical output for all commands (XCP data will be handled separately, eg. violation status)
if not args.violation:
self.logger.store_output(args.gpu, 'values', monitor_values)
# Now handling the single gpu case only
if multiple_devices:
self.logger.store_multiple_device_output()
return
if watching_output and not self.logger.destination == "stdout": # End of single gpu add to watch_output
self.logger.store_watch_output(multiple_device_enabled=False)
if args.violation:
# Print violation status for single gpu, which have different xcp information per 1 gpu
self.logger.print_output(multiple_device_enabled=True, watching_output=watching_output, tabular=True, dual_csv_output=dual_csv_output)
else:
# Print the output for single gpu, which currently does not have multiple xcp information
self.logger.print_output(multiple_device_enabled=False, watching_output=watching_output, tabular=True, dual_csv_output=dual_csv_output)
def xgmi(self, args, multiple_devices=False, gpu=None, metric=None, xgmi_source_status=None, xgmi_link_status=None):
""" Get topology information for target gpus
params:
args - argparser args to pass to subcommand
multiple_devices (bool) - True if checking for multiple devices
gpu (device_handle) - device_handle for target device
metric (bool) - Value override for args.metric
xgmi_source_status (bool) - Value override for args.xgmi_source_status
xgmi_link_status (bool) - Value override for args.xgmi_link_status
return:
Nothing
"""
# Not supported with partitions
# Set args.* to passed in arguments
if gpu:
args.gpu = gpu
if metric:
args.metric = metric
if xgmi_link_status:
args.link_status = xgmi_link_status
if xgmi_source_status:
args.source_status = xgmi_source_status
# Handle No GPU passed
if args.gpu == None:
args.gpu = self.device_handles
if not isinstance(args.gpu, list):
args.gpu = [args.gpu]
# Handle all args being false
if not any([args.metric, args.link_status, args.source_status]):
args.metric = True
args.link_status = True
args.source_status = True
# Clear the table header
self.logger.table_header = ''.rjust(7)
if not self.group_check_printed:
self.helpers.check_required_groups(check_render=True, check_video=False)
self.group_check_printed = True
# Populate the possible gpus and their bdfs
xgmi_values = []
for gpu in args.gpu:
primary_partition = self.helpers.is_primary_partition(gpu)
if not primary_partition:
logging.debug(f"Skipping xgmi command due to non zero partition {gpu}")
continue
logging.debug("check1 device_handle: %s", gpu)
gpu_id = self.helpers.get_gpu_id_from_device_handle(gpu)
gpu_bdf = amdsmi_interface.amdsmi_get_gpu_device_bdf(gpu)
xgmi_values.append({"gpu" : gpu_id,
"bdf" : gpu_bdf})
# Populate header with just it's gpu_id
self.logger.table_header += f"GPU{gpu_id}".rjust(13)
# Cache processor handles for each BDF
src_gpu_handles = {}
for dict in xgmi_values:
try:
src_gpu_handles[dict['bdf']] = amdsmi_interface.amdsmi_get_processor_handle_from_bdf(dict['bdf'])
except amdsmi_exception.AmdSmiLibraryException as e:
logging.debug("Failed to get processor handle for %s | %s", dict['bdf'], e.get_error_info())
src_gpu_handles[dict['bdf']] = None
if args.metric:
# prepend link metrics header to the table header
link_metrics_header = " " + "bdf".ljust(14) + \
"bit_rate".ljust(10) + "max_bandwidth".ljust(15) + \
"link_type".ljust(11)
self.logger.table_header = link_metrics_header + self.logger.table_header.strip()
# Populate dictionary according to format
for xgmi_dict in xgmi_values:
src_gpu_id = xgmi_dict['gpu']
src_gpu_bdf = xgmi_dict['bdf']
src_gpu = src_gpu_handles.get(src_gpu_bdf)
logging.debug("check2 device_handle: %s", src_gpu)
# This should be the same order as the check1
xgmi_dict['link_metrics'] = {
"bit_rate" : "N/A",
"max_bandwidth" : "N/A",
"link_type" : "N/A",
"links" : []
}
xgmi_metrics_info = {"links": []}
try:
xgmi_metrics_info = amdsmi_interface.amdsmi_get_link_metrics(src_gpu)
bitrate = xgmi_metrics_info['links'][0]['bit_rate']
max_bandwidth = xgmi_metrics_info['links'][0]['max_bandwidth']
except amdsmi_exception.AmdSmiLibraryException as e:
bitrate = "N/A"
max_bandwidth = "N/A"
logging.debug("Failed to get bitrate and bandwidth for GPU %s | %s", src_gpu_id,
e.get_error_info())
# Populate bitrate and max_bandwidth with units logic
bw_unit = 'Gb/s'
if self.logger.is_human_readable_format():
xgmi_dict['link_metrics']['bit_rate'] = f"{bitrate} {bw_unit}"
xgmi_dict['link_metrics']['max_bandwidth'] = f"{max_bandwidth} {bw_unit}"
elif self.logger.is_json_format():
xgmi_dict['link_metrics']['bit_rate'] = {"value" : bitrate,
"unit" : bw_unit}
xgmi_dict['link_metrics']['max_bandwidth'] = {"value" : max_bandwidth,
"unit" : bw_unit}
elif self.logger.is_csv_format():
xgmi_dict['link_metrics']['bit_rate'] = bitrate
xgmi_dict['link_metrics']['max_bandwidth'] = max_bandwidth
# Populate link metrics
for dest_gpu in args.gpu:
primary_partition = self.helpers.is_primary_partition(dest_gpu)
if not primary_partition:
continue
dest_gpu_id = self.helpers.get_gpu_id_from_device_handle(dest_gpu)
dest_gpu_bdf = amdsmi_interface.amdsmi_get_gpu_device_bdf(dest_gpu)
dest_link_dict = {
"gpu" : dest_gpu_id,
"bdf" : dest_gpu_bdf,
"read" : 0,
"write" : 0,
}
found = False
for link in xgmi_metrics_info['links']:
if link['bdf'] == dest_gpu_bdf:
# Accumulate read/write if multiple links have the same bdf
dest_link_dict['read'] += link['read']
dest_link_dict['write'] += link['write']
found = True
if not found:
dest_link_dict['read'] = "N/A"
dest_link_dict['write'] = "N/A"
else:
data_unit = 'KB'
if self.logger.is_human_readable_format():
dest_link_dict['read'] = self.helpers.convert_bytes_to_readable(dest_link_dict['read'] * 1024, True)
dest_link_dict['write'] = self.helpers.convert_bytes_to_readable(dest_link_dict['write'] * 1024, True)
elif self.logger.is_json_format():
dest_link_dict['read'] = {"value" : dest_link_dict['read'],
"unit" : data_unit}
dest_link_dict['write'] = {"value" : dest_link_dict['write'],
"unit" : data_unit}
try:
link_type = amdsmi_interface.amdsmi_topo_get_link_type(src_gpu, dest_gpu)['type']
if xgmi_dict['link_metrics']['link_type'] != "XGMI" and isinstance(link_type, int):
if link_type == amdsmi_interface.amdsmi_wrapper.AMDSMI_LINK_TYPE_INTERNAL:
xgmi_dict['link_metrics']['link_type'] = "UNKNOWN"
elif link_type == amdsmi_interface.amdsmi_wrapper.AMDSMI_LINK_TYPE_PCIE:
xgmi_dict['link_metrics']['link_type'] = "PCIE"
elif link_type == amdsmi_interface.amdsmi_wrapper.AMDSMI_LINK_TYPE_XGMI:
xgmi_dict['link_metrics']['link_type'] = "XGMI"
except amdsmi_exception.AmdSmiLibraryException as e:
logging.debug("Failed to get link type for %s to %s | %s",
self.helpers.get_gpu_id_from_device_handle(src_gpu),
self.helpers.get_gpu_id_from_device_handle(dest_gpu),
e.get_error_info())
xgmi_dict['link_metrics']['links'].append(dest_link_dict)
# Handle printing for tabular format
if self.logger.is_human_readable_format():
# Populate tabular output
tabular_output = []
for xgmi_dict in xgmi_values:
tabular_output_dict = {}
# Create GPU row and add to tabular_output
for key, value in xgmi_dict.items():
if key == "gpu":
tabular_output_dict["gpu#"] = f"GPU{value}"
if key == "bdf":
tabular_output_dict["bdf"] = value
if key == "link_metrics":
for link_key, link_value in value.items():
if link_key == "bit_rate":
tabular_output_dict["bit_rate"] = link_value
if link_key == "max_bandwidth":
tabular_output_dict["max_bandwidth"] = link_value
if link_key == "link_type":
tabular_output_dict["link_type"] = link_value
tabular_output.append(tabular_output_dict)
# Create Read and Write rows and add to tabular_output
read_output_dict = {"RW" : " Read"}
write_output_dict = {"RW" : " Write"}
for key, value in xgmi_dict.items():
if key == "link_metrics":
for link_key, link_value in value.items():
if link_key == "links":
for link in link_value:
read_output_dict[f"bdf_{link['gpu']}"] = link["read"]
write_output_dict[f"bdf_{link['gpu']}"] = link["write"]
tabular_output.append(read_output_dict)
tabular_output.append(write_output_dict)
# Print out the tabular output
self.logger.multiple_device_output = tabular_output
self.logger.table_title = "\nLINK METRIC TABLE"
self.logger.print_output(multiple_device_enabled=True, tabular=True)
self.logger.multiple_device_output = xgmi_values
if self.logger.is_csv_format():
new_output = []
for elem in self.logger.multiple_device_output:
new_output.append(self.logger.flatten_dict(elem, topology_override=True))
self.logger.multiple_device_output = new_output
if self.logger.is_json_format():
self.logger.store_xgmi_metric_json_output.append(xgmi_values)
if not any([args.link_status, args.source_status]):
self.logger.combine_arrays_to_json()
elif not self.logger.is_human_readable_format():
self.logger.print_output(multiple_device_enabled=True)
if args.source_status:
# Header modification
self.logger.table_header = ''.rjust(7)
current_header = " ".ljust(7) + \
"bdf".ljust(14) + \
"port_num".ljust(20)
self.logger.table_header = current_header + self.logger.table_header.strip()
# Process each GPU
tabular_output = []
for xgmi_dict in xgmi_values:
src_gpu_id = xgmi_dict['gpu']
src_gpu_bdf = xgmi_dict['bdf']
src_gpu = src_gpu_handles.get(src_gpu_bdf)
# Populate link statuses
tabular_output_dict = {"gpu#": f"GPU{src_gpu_id}",
"gpu": src_gpu_id,
"bdf": src_gpu_bdf,
"link_status": "N/A"}
try:
link_status = amdsmi_interface.amdsmi_get_gpu_xgmi_link_status(src_gpu)
tabular_output_dict['link_status'] = link_status['status']
if self.logger.is_human_readable_format():
del tabular_output_dict['gpu']
else:
del tabular_output_dict['gpu#']
tabular_output.append(tabular_output_dict)
if self.logger.is_json_format():
self.logger.store_xgmi_source_status_json_output.append(tabular_output_dict)
except amdsmi_exception.AmdSmiLibraryException as e:
xgmi_dict['link_metrics']['link_status']={"status": "failed"}
logging.debug("Failed to get XGMI link status for GPU %s | %s", src_gpu_id, e.get_error_info())
#populate link status data for output
if self.logger.is_human_readable_format():
xgmi_dict['link_status'] = tabular_output
self.logger.multiple_device_output= tabular_output
self.logger.table_title = "\nGPU LINK PORT STATUS"
if not self.logger.is_json_format():
self.logger.print_output(multiple_device_enabled=True, tabular=True)
self.logger.clear_multiple_devices_output()
if self.logger.is_json_format():
if not args.link_status:
self.logger.combine_arrays_to_json()
if args.link_status:
# XGMI LINK STATUS for src_gpu to dest_gpu
header = [" ".ljust(8), "bdf".ljust(15)] + [f"GPU{d['gpu']}".ljust(14) for d in xgmi_values]
self.logger.table_header = "".join(header)
self.logger.table_title = "\nXGMI LINK STATUS"
src_link_status_map = {}
for gpu_dict in xgmi_values:
src_gpu_id = gpu_dict['gpu']
src_gpu_bdf = gpu_dict['bdf']
src_gpu = src_gpu_handles.get(src_gpu_bdf)
try:
link_status = amdsmi_interface.amdsmi_get_gpu_xgmi_link_status(src_gpu)
src_link_status_map[src_gpu_bdf] = link_status['status']
except amdsmi_exception.AmdSmiLibraryException:
src_link_status_map[src_gpu_bdf] = ["N/A"] * amdsmi_interface.AMDSMI_MAX_NUM_XGMI_LINKS
tabular_output = []
for src_xgmi_dict in xgmi_values:
src_gpu_id = src_xgmi_dict['gpu']
src_gpu_bdf = src_xgmi_dict['bdf']
src_gpu = src_gpu_handles.get(src_gpu_bdf)
try:
xgmi_metrics_info = amdsmi_interface.amdsmi_get_link_metrics(src_gpu)
except amdsmi_exception.AmdSmiLibraryException:
xgmi_metrics_info = {"links": []}
# First column: GPU# + tab + bdf, then status for each dest bdf
if self.logger.is_human_readable_format():
row_dict = {"": f"GPU{src_gpu_id}\t{src_gpu_bdf}".ljust(20)}
else:
row_dict = {"gpu": f"GPU{src_gpu_id}", "bdf": src_gpu_bdf}
json_status = []
# Cache GPU handles for destination GPUs
dest_gpu_handles = {dest_xgmi_dict['bdf']:
amdsmi_interface.amdsmi_get_processor_handle_from_bdf(dest_xgmi_dict['bdf'])
for dest_xgmi_dict in xgmi_values}
for dest_xgmi_dict in xgmi_values:
dest_gpu_bdf = dest_xgmi_dict['bdf']
dest_gpu = dest_gpu_handles[dest_gpu_bdf]
# Find all link indexes in xgmi_metrics_info for this destination
link_indexes = []
for idx, link in enumerate(xgmi_metrics_info['links']):
if link['bdf'] == dest_gpu_bdf:
link_indexes.append(idx)
# Use the found link index to get the status if valid
if link_indexes and len(link_indexes) <= len(src_link_status_map.get(src_gpu_bdf, [])):
statuses = []
for link_idx in link_indexes:
if link_idx < len(src_link_status_map[src_gpu_bdf]):
statuses.append(str(src_link_status_map[src_gpu_bdf][link_idx]))
# Join multiple statuses with "/"
if statuses:
status = "/".join(statuses)
else:
status = "N/A"
elif dest_gpu_bdf == src_gpu_bdf:
status = "SELF"
else:
status = "N/A"
if self.logger.is_human_readable_format():
row_dict[dest_gpu_bdf.ljust(14)] = str(status).ljust(14)
else:
row_dict[dest_gpu_bdf] = status
json_status.append(status)
tabular_output.append(row_dict)
if self.logger.is_json_format():
self.logger.store_xgmi_link_status_json_output.append({
"gpu": src_gpu_id,
"bdf": src_gpu_bdf,
"link_status": json_status
})
if not self.logger.is_json_format():
self.logger.multiple_device_output = tabular_output
self.logger.print_output(multiple_device_enabled=True, tabular=True)
self.logger.clear_multiple_devices_output()
if self.logger.is_json_format():
self.logger.combine_arrays_to_json()
if self.logger.is_human_readable_format():
# Populate the legend output
legend_parts = [
"\n\nLegend:",
" SELF = Current GPU",
" N/A = Not supported",
" U / D / X = Link is Up / Down / Disabled",
" Read / Write = GPU Metric Accumulated Read / Write"
]
legend_output = "\n".join(legend_parts)
if self.logger.destination == 'stdout':
print(legend_output)
else:
with self.logger.destination.open('a', encoding="utf-8") as output_file:
output_file.write(legend_output + '\n')
def partition(self, args, multiple_devices=False, gpu=None, current=None, memory=None, accelerator=None):
""" Display parition information for the target GPU
param:
args - argparser args to pass to subcommand
multiple_devices (bool) - True if checking for multiple devices
gpu (device_handle) - device_handle for target device
current - boolean which dictates whether the current partition information is shown
memory - boolean which dictates whether the memory partition information is shown
accelerator - boolean which dictates whether the accelerator partition information is shown
returns:
nothing
"""
if gpu:
args.gpu = gpu
if args.gpu == None:
args.gpu = self.device_handles
if not isinstance(args.gpu, list):
args.gpu = [args.gpu]
if current:
args.current = current
if memory:
args.memory = memory
if accelerator:
args.accelerator = accelerator
if not self.group_check_printed:
self.helpers.check_required_groups(check_render=True, check_video=False)
self.group_check_printed = True
###########################################
# amd-smi partition (no args) #
###########################################
# if no args are present, then everything should be displayed
if not args.current and not args.memory and not args.accelerator:
args.current = True
args.memory = True
args.accelerator = True
###########################################
# amd-smi partition --current #
###########################################
if args.current:
self.logger.table_header = ''.rjust(7)
current_header = "GPU_ID".ljust(8) + \
"MEMORY".ljust(8) + \
"ACCELERATOR_TYPE".ljust(18) + \
"ACCELERATOR_PROFILE_INDEX".ljust(27) + \
"PARTITION_ID".ljust(14)
self.logger.table_header = current_header + self.logger.table_header.strip()
tabular_output = []
for gpu in args.gpu:
gpu_id = self.helpers.get_gpu_id_from_device_handle(gpu)
try:
partition_dict = amdsmi_interface.amdsmi_get_gpu_accelerator_partition_profile(gpu)
partition_id = str(partition_dict['partition_id']).replace("[", "").replace("]", "").replace(" ", "")
profile_type = partition_dict['partition_profile']['profile_type']
profile_index = partition_dict['partition_profile']['profile_index']
except amdsmi_exception.AmdSmiLibraryException as e:
profile_type = "N/A"
profile_index = "N/A"
partition_id = str(partition_dict['partition_id']).replace("[", "").replace("]", "").replace(" ", "")
logging.debug("Failed to get accelerator partition profile for GPU %s | %s", gpu_id, e.get_error_info())
try:
current_mem_cap = amdsmi_interface.amdsmi_get_gpu_memory_partition(gpu)
except amdsmi_exception.AmdSmiLibraryException as e:
current_mem_cap = "N/A"
logging.debug("Failed to get current memory partition capabilties for GPU %s | %s", gpu_id, e.get_error_info())
if profile_type == 0:
profile_type = "N/A"
tabular_output_dict = {"gpu_id": gpu_id,
"memory": current_mem_cap,
"accelerator_type": profile_type,
"accelerator_profile_index": profile_index,
"partition_id": partition_id}
tabular_output.append(tabular_output_dict)
self.logger.multiple_device_output = tabular_output
self.logger.table_title = "CURRENT_PARTITION"
if self.logger.is_json_format():
self.logger.store_current_partition_json_output.extend(tabular_output)
if not (args.memory or args.accelerator):
self.logger.combine_arrays_to_json()
else:
self.logger.print_output(multiple_device_enabled=True, tabular=True, dynamic=True)
self.logger.clear_multiple_devices_output()
###########################################
# amd-smi partition --memory #
###########################################
if args.memory:
tabular_output = []
self.logger.table_header = ''.rjust(7)
current_header = "GPU_ID".ljust(8) + \
"MEMORY_PARTITION_CAPS".ljust(23) + \
"CURRENT_MEMORY_PARTITION".ljust(26)
self.logger.table_header = current_header + self.logger.table_header.strip()
for gpu in args.gpu:
gpu_id = self.helpers.get_gpu_id_from_device_handle(gpu)
mem_caps_str = "N/A"
current_memory_partition = "N/A"
try:
memory_partition_config = amdsmi_interface.amdsmi_get_gpu_memory_partition_config(gpu)
mem_caps_str = str(memory_partition_config['partition_caps']).replace("]", "").replace("[", "").replace("\'", "").replace(" ", "")
current_memory_partition = memory_partition_config['mp_mode']
except amdsmi_exception.AmdSmiLibraryException as e:
logging.debug("Failed to get current memory partition for GPU %s | %s", gpu_id, e.get_error_info())
tabular_output_dict = {"gpu_id": gpu_id,
"memory_partition_caps": mem_caps_str,
"current_memory_partition": current_memory_partition}
tabular_output.append(tabular_output_dict)
self.logger.multiple_device_output = tabular_output
self.logger.table_title = "\nMEMORY_PARTITION"
if self.logger.is_json_format():
self.logger.store_memory_partition_json_output.extend(tabular_output)
if not args.accelerator:
self.logger.combine_arrays_to_json()
else:
self.logger.print_output(multiple_device_enabled=True, tabular=True, dynamic=True)
self.logger.clear_multiple_devices_output()
###########################################
# amd-smi partition --accelerator #
###########################################
if args.accelerator:
self.logger.table_header = ''.rjust(7)
current_header = "GPU_ID".ljust(8) + \
"PROFILE_INDEX".ljust(15) + \
"MEMORY_PARTITION_CAPS".ljust(23) + \
"ACCELERATOR_TYPE".ljust(18) + \
"PARTITION_ID".ljust(17) + \
"NUM_PARTITIONS".ljust(16) + \
"NUM_RESOURCES".ljust(15) + \
"RESOURCE_INDEX".ljust(16) + \
"RESOURCE_TYPE".ljust(15) + \
"RESOURCE_INSTANCES".ljust(20) + \
"RESOURCES_SHARED".ljust(18)
self.logger.table_header = current_header + self.logger.table_header.strip()
tabular_output = []
prev_gpu_id = "N/A"
for gpu in args.gpu:
gpu_id = self.helpers.get_gpu_id_from_device_handle(gpu)
tabular_output_dict = {"gpu_id": gpu_id,
"profile_index": "N/A",
"memory_partition_caps": "N/A",
"accelerator_type": "N/A",
"partition_id": "0",
"num_partitions": "N/A",
"num_resources": "N/A",
"resource_index": "N/A",
"resource_type": "N/A",
"resource_instances": "N/A",
"resources_shared": "N/A"}
try:
partition_dict = amdsmi_interface.amdsmi_get_gpu_accelerator_partition_profile(gpu)
partition_id = str(partition_dict['partition_id']).replace("[", "").replace("]", "").replace(" ", "")
current_accelerator_type = partition_dict['partition_profile']['profile_type']
tabular_output_dict["partition_id"] = partition_id
# save only the primary GPU node's partition_id (the 1st listed device; non N/A one)
# else keep current_partition_id unchanged for displaying in accelerator resource's output
if partition_id != "N/A":
current_partition_id = partition_id
except amdsmi_exception.AmdSmiLibraryException as e:
profile_type = "N/A"
profile_index = "N/A"
partition_id = "0"
mem_caps_str = "N/A"
num_partitions = 0
current_accelerator_type = "N/A"
logging.debug("Failed to get accelerator partition profile for GPU %s | %s", gpu_id, e.get_error_info())
try:
partition_config_dict = amdsmi_interface.amdsmi_get_gpu_accelerator_partition_profile_config(gpu)
logging.debug("amdsmi_commands.py | partition_config_dict: " + str(json.dumps(partition_config_dict, indent=4)))
num_profiles = partition_config_dict['num_profiles']
num_resource_profiles = partition_config_dict['num_resource_profiles']
resource_index = 0
prev_accelerator_type = "N/A"
for p in range(0, num_profiles):
accelerator_type = partition_config_dict['profiles'][p]['profile_type']
profile_index = partition_config_dict['profiles'][p]['profile_index']
num_partitions = partition_config_dict['profiles'][p]['num_partitions']
mem_caps_str = str(partition_config_dict['profiles'][p]['memory_caps']).replace("]", "").replace("[", "").replace("\'", "").replace(" ", "")
# 2 modifications based on the current accelerator type:
# 1) display a * for the current accelerator type, otherwise display as normal
# 2) display partition id only for the current accelerator profile (the *'d one)
if current_accelerator_type == accelerator_type:
accelerator_type = accelerator_type + "*"
partition_id = current_partition_id
else:
partition_id = "N/A"
# only display the first instance of the gpu_id, rest are empty strings
if prev_gpu_id != gpu_id:
tabular_gpu_id = gpu_id
prev_gpu_id = gpu_id
else:
tabular_gpu_id = ""
logging.debug("amdsmi_commands.py | tabular_gpu_id: " + str(tabular_gpu_id))
if num_resource_profiles == 0:
if prev_accelerator_type != accelerator_type: # only print the first instance of the resources
tabular_output_dict = {"gpu_id": tabular_gpu_id,
"profile_index": profile_index,
"memory_partition_caps": mem_caps_str,
"accelerator_type": accelerator_type,
"partition_id": partition_id,
"num_partitions": num_partitions,
"num_resources": num_resource_profiles,
"resource_index": "N/A",
"resource_type": "N/A",
"resource_instances": "N/A",
"resources_shared": "N/A"}
prev_accelerator_type = accelerator_type
tabular_output.append(tabular_output_dict)
continue
for r in range(0, num_resource_profiles):
logging.debug("amdsmi_commands.py | p: " + str(p) + "; r: " + str(r)
+ "; accelerator_type: " + str(accelerator_type))
resource_type = partition_config_dict['profiles'][p]['resources'][r]['resource_type']
resource_instances = partition_config_dict['profiles'][p]['resources'][r]['partition_resource']
resources_shared = partition_config_dict['profiles'][p]['resources'][r]['num_partitions_share_resource']
if prev_accelerator_type != accelerator_type: # only print the first instance of the resources
tabular_output_dict = {"gpu_id": tabular_gpu_id,
"profile_index": profile_index,
"memory_partition_caps": mem_caps_str,
"accelerator_type": accelerator_type,
"partition_id": partition_id,
"num_partitions": num_partitions,
"num_resources": num_resource_profiles,
"resource_index": resource_index,
"resource_type": resource_type,
"resource_instances": resource_instances,
"resources_shared": resources_shared}
prev_accelerator_type = accelerator_type
else:
tabular_output_dict = {"gpu_id": "",
"profile_index": "",
"memory_partition_caps": "",
"accelerator_type": "",
"partition_id": "",
"num_partitions": "",
"num_resources": "",
"resource_index": resource_index,
"resource_type": resource_type,
"resource_instances": resource_instances,
"resources_shared": resources_shared}
resource_index += 1
tabular_output.append(tabular_output_dict)
except amdsmi_exception.AmdSmiLibraryException as e:
tabular_output.append(tabular_output_dict)
self.logger.multiple_device_output = tabular_output
self.logger.table_title = "\nACCELERATOR_PARTITION_PROFILES"
# only display warning message if not running as root or with sudo
if os.geteuid() != 0:
self.logger.warning_message = """
***************************************************************************
** WARNING: **
** ACCELERATOR_PARTITION_PROFILES requires sudo/root permissions to run. **
** Please run the command with sudo permissions to get accurate results. **
***************************************************************************
"""
if self.logger.is_json_format():
self.logger.store_partition_profiles_json_output.extend(tabular_output)
else:
self.logger.print_output(multiple_device_enabled=True, tabular=True, dynamic=True)
self.logger.clear_multiple_devices_output()
self.logger.warning_message = "" # clear the warning message
#########################################
# print accelerator partition resources #
#########################################
self.logger.table_header = ''.rjust(7)
current_header = "RESOURCE_INDEX".ljust(16) + \
"RESOURCE_TYPE".ljust(15) + \
"RESOURCE_INSTANCES".ljust(20) + \
"RESOURCES_SHARED".ljust(18)
self.logger.table_header = current_header + self.logger.table_header.strip()
tabular_output = []
for gpu in args.gpu:
gpu_id = self.helpers.get_gpu_id_from_device_handle(gpu)
tabular_output_dict = {"resource_index": "N/A",
"resource_type": "N/A",
"resource_instances": "N/A",
"resources_shared": "N/A"}
try:
partition_config_dict = amdsmi_interface.amdsmi_get_gpu_accelerator_partition_profile_config(gpu)
logging.debug("amdsmi_commands.py | partition_config_dict: " + str(json.dumps(partition_config_dict, indent=4)))
num_profiles = partition_config_dict['num_profiles']
num_resource_profiles = partition_config_dict['num_resource_profiles']
if num_resource_profiles == 0:
tabular_output.append(tabular_output_dict)
continue
resource_index = 0
for p in range(0, num_profiles):
for r in range(0, num_resource_profiles):
resource_type = partition_config_dict['profiles'][p]['resources'][r]['resource_type']
resource_instances = partition_config_dict['profiles'][p]['resources'][r]['partition_resource']
resources_shared = partition_config_dict['profiles'][p]['resources'][r]['num_partitions_share_resource']
tabular_output_dict = {
"resource_index": resource_index,
"resource_type": resource_type,
"resource_instances": resource_instances,
"resources_shared": resources_shared}
resource_index += 1
tabular_output.append(tabular_output_dict)
except amdsmi_exception.AmdSmiLibraryException as e:
tabular_output.append(tabular_output_dict)
self.logger.multiple_device_output = tabular_output
self.logger.table_title = "\nACCELERATOR_PARTITION_RESOURCES"
if self.logger.is_json_format():
self.logger.store_partition_resources_json_output.extend(tabular_output)
else:
self.logger.print_output(multiple_device_enabled=True, tabular=True, dynamic=True)
if self.logger.is_json_format():
self.logger.combine_arrays_to_json()
self.logger.clear_multiple_devices_output()
if self.logger.is_human_readable_format():
# print legend
legend_parts = [
"\n\nLegend:",
" * = Current mode"]
legend_output = "\n".join(legend_parts)
if self.logger.destination == 'stdout':
print(legend_output)
else:
with self.logger.destination.open('a', encoding="utf-8") as output_file:
output_file.write(legend_output + '\n')
def ras(self, args, multiple_devices=False, gpu=None, cper=None, afid=None,
severity=None, folder=None, file_limit=None, cper_file=None, follow=None):
"""
Retrieve and process CPER (RAS) entries for a target GPU.
Expected command (all options only):
amd-smi ras --cper --severity=nonfatal-uncorrected,fatal --folder <folder_name> --file-limit=1000 --follow
Since no timestamp is provided on the command line, the function starts from a default cursor of 0.
The output file name is auto-generated using the timestamp from the CPER header data (converted from
the header’s "YYYY/MM/DD HH:MM:SS" format), along with the GPU/platform ID and error severity.
"""
# GPU handle logic.
if gpu:
args.gpu = gpu
if cper:
args.cper = cper
if afid:
args.afid = afid
if severity:
args.severity = severity
if folder:
args.folder = folder
if file_limit:
args.file_limit = file_limit
if cper_file:
args.cper_file = cper_file
if follow:
args.follow = follow
if args.gpu == None:
args.gpu = self.device_handles
if args.afid:
if args.cper_file:
afids = self.helpers.pvtDumpAfids(args.cper_file)
print(' '.join(map(str, afids)))
return
else:
command = " ".join(sys.argv[1:])
message = f"Command '{command}' requires '--cper-file'. Run '--help' for more info."
raise AmdSmiInvalidCommandException(command,
self.logger.format,
message)
if not self.group_check_printed:
self.helpers.check_required_groups(check_render=True, check_video=True)
self.group_check_printed = True
if not args.cper:
return
if not args.gpu:
return
if not isinstance(args.gpu, list):
args.gpu = [args.gpu]
args.cursor = [0] * len(args.gpu)
# Using all the devices given in args.gpu
# Populate a list of all the primary partition GPU ids (GPU 0, GPU 1, etc)
partition_warning_flag = True
primary_partition_gpu_ids = set() # set of all primary partition GPU ids from arg.gpu
for device_handle in args.gpu:
# First get the partition
partition_id = self.helpers.get_partition_id(device_handle)
# If there is a single primary partition within args.gpu then we don't need to print the warning
if partition_id == 0:
partition_warning_flag = False
break
# Then attempt to get the primary GPU id for that partition
primary_partition_gpu_id = self.helpers.get_primary_partition_gpu_id(device_handle)
# Add to the set if it's a non-primary partition and we found a valid primary GPU id
if partition_id != 0 and primary_partition_gpu_id is not None:
primary_partition_gpu_ids.add(primary_partition_gpu_id)
if partition_warning_flag:
# Create a list of the primary partitions
primary_partitions_str = " ".join(f"GPU{gpu_id}" for gpu_id in primary_partition_gpu_ids)
print("WARNING: CPER files are only available on primary partitions")
if len(primary_partition_gpu_ids) > 1:
print(f"Try with primary partitions {primary_partitions_str}",end="")
else:
print(f"Try with primary partition {primary_partitions_str}",end="")
print()
while True:
for idx, device_handle in enumerate(args.gpu):
self.helpers.ras_cper(args, device_handle, self.logger, idx)
if not args.follow:
break
time.sleep(1)
def node(self, args, multiple_devices=False, nodes=None, power_management=None):
"""List node informations
Args:
args (Namespace): Namespace containing the parsed CLI args
multiple_devices (bool, optional): True if checking for multiple devices.
Defaults to False.
Returns:
None: Print output via AMDSMILogger to destination
"""
# Set args.* to passed in arguments
if nodes:
args.nodes = nodes
if power_management:
args.power_management = power_management
if getattr(args, 'nodes', None) is None:
args.nodes = self.node_handle
if not self.group_check_printed:
self.helpers.check_required_groups(check_render=True, check_video=False)
self.group_check_printed = True
# Get NPM info
if args.nodes is not None:
try:
npm_info = amdsmi_interface.amdsmi_get_npm_info(args.nodes)
except amdsmi_exception.AmdSmiLibraryException as e:
logging.debug("amdsmi_get_npm_info failed: %s", e.get_error_info())
npm_info = "N/A"
else:
logging.debug('No node handle available to query NPM info')
npm_info = "N/A"
# Log outputs
npm_dict = {"limit": "N/A", "status": "N/A"}
power_unit ="W"
limit = "N/A"
if isinstance(npm_info, dict):
limit = npm_info.get('limit', "N/A")
status = npm_info.get('status', npm_info.get('current', "N/A"))
if limit !="N/A":
npm_dict['limit'] = limit
status = "DISABLED" if status == amdsmi_interface.amdsmi_wrapper.AMDSMI_NPM_STATUS_DISABLED else "ENABLED"
npm_dict.update({"status": status})
if self.logger.is_human_readable_format() and self.logger.destination == 'stdout':
print(f"NODE:\n POWER_MANAGEMENT:\n LIMIT: {npm_dict.get('limit', 'N/A')} {power_unit}\n STATUS: {npm_dict.get('status', 'N/A')}")
else:
if self.logger.is_csv_format():
csv_dict = {}
csv_dict['limit'] = npm_dict.get('limit', "N/A")
csv_dict['status'] = npm_dict.get('status', "N/A")
self.logger.output = csv_dict
else:
# For JSON and human readable format with file output
npm_dict["limit"] = self.helpers.unit_format(self.logger, limit, power_unit)
self.logger.output = {'node': {'power_management': npm_dict}}
if multiple_devices:
self.logger.store_multiple_device_output()
return
self.logger.print_output()
def default(self, args):
"""Display the default amdsmi view when no args are given."""
# check groups first
if not self.group_check_printed:
self.helpers.check_required_groups(check_render=True, check_video=False)
self.group_check_printed = True
processors = amdsmi_interface.amdsmi_get_processor_handles()
version_info = {"amd-smi": "N/A",
"amdgpu version": "N/A",
"kernel version": "N/A",
"fw pldm version": "N/A",
"vbios version": "N/A",
"rocm version": (False, "N/A")}
version_info['rocm version'] = amdsmi_interface.amdsmi_get_rocm_version()
version_info['kernel version'] = os.uname().release
try:
version_info["amdgpu version"] = amdsmi_interface.amdsmi_get_gpu_driver_info(processors[0])
except amdsmi_exception.AmdSmiLibraryException as e:
version_info["amdgpu version"] = "N/A"
logging.debug("Failed to get driver info for gpu: %s", e.get_error_info())
try:
fw_info = amdsmi_interface.amdsmi_get_fw_info(processors[0])
for fw in fw_info['fw_list']:
if "pldm" in fw.keys():
version_info['fw pldm version'] = fw['pldm']
# we only need to find one of them
break
except amdsmi_exception.AmdSmiLibraryException as e:
version_info['fw pldm version'] = "N/A"
logging.debug("Failed to get fw pldm info for gpu: %s", e.get_error_info())
try:
version_info['vbios version'] = amdsmi_interface.amdsmi_get_gpu_vbios_info(processors[0])["version"]
if version_info['vbios version'] == "":
version_info['vbios version'] = "N/A"
except amdsmi_exception.AmdSmiLibraryException as e:
version_info['vbios version'] = "N/A"
logging.debug("Failed to get vbios info for gpu: %s", e.get_error_info())
version_info["amd-smi"] = f'{__version__}'
default_table_info_dict = {}
default_table_info_dict.update({"version_info": version_info})
gpu_info_list = []
all_process_list = []
# get info for each processor to display in default output
for processor in processors:
gpu_info_dict = {}
gpu_id = self.helpers.get_gpu_id_from_device_handle(processor)
gpu_info_dict.update({"gpu_id": gpu_id})
# get common gpu_metrics first
try:
gpu_metrics = amdsmi_interface.amdsmi_get_gpu_metrics_info(processor)
except amdsmi_exception.AmdSmiLibraryException as e:
gpu_metrics = amdsmi_interface._NA_amdsmi_get_gpu_metrics_info()
# partition info
try:
current_mem = amdsmi_interface.amdsmi_get_gpu_memory_partition(processor)
except amdsmi_exception.AmdSmiLibraryException as e:
current_mem = "N/A"
try:
current_comp = amdsmi_interface.amdsmi_get_gpu_compute_partition(processor)
except amdsmi_exception.AmdSmiLibraryException as e:
current_comp = "N/A"
if current_comp == "N/A" or current_mem == "N/A":
partition_mode = "N/A"
else:
partition_mode = f"{current_comp}/{current_mem}"
gpu_info_dict.update({"partition_mode": partition_mode})
# GPU name market name and OAM ID
try:
asic_info = amdsmi_interface.amdsmi_get_gpu_asic_info(processor)
market_name = asic_info['market_name']
oam_id = asic_info['oam_id']
# get num_cu now for use later
total_num_cu = float(asic_info['num_compute_units'])
except amdsmi_exception.AmdSmiLibraryException as e:
market_name = "N/A"
oam_id = "N/A"
total_num_cu = "N/A"
gpu_info_dict.update({"market_name": market_name})
gpu_info_dict.update({"oam_id": oam_id})
# bdf
try:
bdf = amdsmi_interface.amdsmi_get_gpu_device_bdf(processor)
# if the len of the bdf is not 12, then invalid values are being populated.
if len(bdf) != 12:
bdf = "N/A"
except amdsmi_exception.AmdSmiLibraryException as e:
bdf = "N/A"
gpu_info_dict.update({"bdf": bdf})
# HIP ID
try:
enum_info = amdsmi_interface.amdsmi_get_gpu_enumeration_info(processor)
hip_id = enum_info['hip_id']
except amdsmi_exception.AmdSmiLibraryException as e:
hip_id = "N/A"
gpu_info_dict.update({"hip_id": hip_id})
# mem utilization, GPU utilization, power usage, and temperature from gpu_metrics
if gpu_metrics != "N/A":
mem_util = gpu_metrics['average_umc_activity']
gfx_util = gpu_metrics['average_gfx_activity']
if gpu_metrics['current_socket_power'] != "N/A":
current_power = gpu_metrics['current_socket_power']
else:
current_power = gpu_metrics['average_socket_power']
# If the hotspot temperature is not available use the edge temp (applicable to APUs)
if gpu_metrics['temperature_hotspot'] != "N/A":
temperature = gpu_metrics['temperature_hotspot']
elif gpu_metrics['temperature_edge'] != "N/A":
temperature = gpu_metrics['temperature_edge']
else:
temperature = "N/A"
else:
mem_util = "N/A"
gfx_util = "N/A"
current_power = "N/A"
temperature = "N/A"
gpu_info_dict.update({"mem_util": mem_util})
gpu_info_dict.update({"gfx_util": gfx_util})
gpu_info_dict.update({"temp": temperature})
# rest of power usage info; Will assume we're always trying to get PPT0 for now
try:
power_cap_info = amdsmi_interface.amdsmi_get_power_cap_info(processor, 0)
socket_power_limit = self.helpers.convert_SI_unit(power_cap_info['power_cap'], AMDSMIHelpers.SI_Unit.MICRO)
power_usage = {"current_power": current_power, "power_limit": socket_power_limit}
except amdsmi_exception.AmdSmiLibraryException as e:
power_usage = "N/A"
gpu_info_dict.update({"power_usage": power_usage})
# memory usage - Use APU-aware memory selection
try:
# Use helper method to determine appropriate memory type
mem_type, mem_type_name = self.helpers.get_apu_memory_type_and_name(processor, gpu_id)
# Get memory usage and total using the determined memory type
used_mem = amdsmi_interface.amdsmi_get_gpu_memory_usage(processor, mem_type) // (1024*1024)
total_mem = amdsmi_interface.amdsmi_get_gpu_memory_total(processor, mem_type) // (1024*1024)
# Create appropriate dictionary keys based on memory type
if mem_type_name == "GTT":
mem_usage = {"used_gtt": used_mem, "total_gtt": total_mem}
else:
mem_usage = {"used_vram": used_mem, "total_vram": total_mem}
except amdsmi_exception.AmdSmiLibraryException as e:
mem_usage = "N/A"
gpu_info_dict.update({"mem_usage": mem_usage})
# uncorrectable ECC errors
try:
ecc_count = amdsmi_interface.amdsmi_get_gpu_total_ecc_count(processor)
uncorrectable = ecc_count.pop('uncorrectable_count')
except amdsmi_exception.AmdSmiLibraryException as e:
uncorrectable = "N/A"
gpu_info_dict.update({"uncorr_ecc": uncorrectable})
# Fan usage
try:
fan_speed = amdsmi_interface.amdsmi_get_gpu_fan_speed(processor, 0)
except amdsmi_exception.AmdSmiLibraryException as e:
logging.debug("Failed to get fan speed for gpu %s | %s", processor, e.get_error_info())
fan_speed = "N/A"
try:
fan_max = amdsmi_interface.amdsmi_get_gpu_fan_speed_max(processor, 0)
fan_usage = "N/A"
if fan_max > 0 and fan_speed != "N/A":
fan_usage = round((float(fan_speed) / float(fan_max)) * 100, 2)
except amdsmi_exception.AmdSmiLibraryException as e:
logging.debug("Failed to get max fan speed for gpu %s | %s", processor, e.get_error_info())
fan_usage = "N/A"
gpu_info_dict.update({"fan": fan_usage})
gpu_info_list.append(gpu_info_dict)
# Running Processes
try:
raw_process_list = amdsmi_interface.amdsmi_get_gpu_process_list(processor)
for proc in raw_process_list:
proc_info_dict = {"gpu": "N/A", "pid": "N/A", "name": "N/A","gtt": "N/A", "vram": "N/A", "mem_usage": "N/A", "cu_occupancy": "N/A"}
proc_info_dict['gpu'] = gpu_id
proc_info_dict['pid'] = proc['pid']
proc_info_dict['name'] = proc['name']
proc_info_dict['gtt'] = self.helpers.convert_bytes_to_readable(proc['memory_usage']['gtt_mem'])
proc_info_dict['vram'] = self.helpers.convert_bytes_to_readable(proc['memory_usage']['vram_mem'])
proc_info_dict['mem_usage'] = self.helpers.convert_bytes_to_readable(proc['mem'])
# Handle cu_occupancy conversion safely
try:
if proc['cu_occupancy'] != "N/A" and total_num_cu != "N/A":
num_cu = float(proc['cu_occupancy'])
proc_info_dict['cu_occupancy'] = {"current_cu": num_cu, "total_num_cu": total_num_cu}
else:
proc_info_dict['cu_occupancy'] = {"current_cu": "N/A", "total_num_cu": total_num_cu}
except (ValueError, TypeError):
proc_info_dict['cu_occupancy'] = {"current_cu": "N/A", "total_num_cu": total_num_cu}
all_process_list.append(proc_info_dict)
except amdsmi_exception.AmdSmiLibraryException as e:
logging.debug("Failed to get process list for gpu %s | %s", gpu_id, e.get_error_info())
default_table_info_dict.update({f"gpu_info_list": gpu_info_list})
default_table_info_dict.update({"processes": all_process_list})
if self.logger.is_json_format():
self.logger.output = default_table_info_dict
self.logger.print_output()
elif self.logger.is_csv_format():
self.logger.multiple_device_output = default_table_info_dict
self.logger.print_output(multiple_device_enabled=True, tabular=True, dynamic=True)
else:
self.logger.print_default_output(default_table_info_dict)
def _event_thread(self, commands, i):
devices = commands.device_handles
if len(devices) == 0:
print("No GPUs on machine")
return
# Check that KFD permissions are available
if not self.group_check_printed:
self.helpers.check_required_groups(check_render=True, check_video=False)
self.group_check_printed = True
device = devices[i]
listener = amdsmi_interface.AmdSmiEventReader(device,
amdsmi_interface.AmdSmiEvtNotificationType)
values_dict = {}
while not self.stop:
try:
events = listener.read(2000)
for event in events:
values_dict["event"] = event["event"]
# parse message as it's own dictionary
message_list = event["message"].split(" ")
message_dict = {}
for item in message_list:
if not item == "":
item_list = item.split(": ")
message_dict.update({item_list[0]: item_list[1]})
values_dict["message"] = message_dict
commands.logger.store_output(event['processor_handle'], 'values', values_dict)
commands.logger.print_output()
except amdsmi_exception.AmdSmiLibraryException as e:
if e.err_code != amdsmi_interface.amdsmi_wrapper.AMDSMI_STATUS_NO_DATA:
print(e)
except Exception as e:
print(e)
listener.stop()
def rocm_smi(self, args):
"""
Display GPU information in ROCm-SMI compatible format (showAllConcise).
This provides a drop-in replacement for rocm-smi --showallconcise using amdsmi backend.
Args:
args: Parsed arguments (unused for this command)
"""
try:
# Import the ROCm-SMI compatible functions from the compatibility module
import sys
import os
# Add the current directory to path if needed
current_dir = os.path.dirname(os.path.abspath(__file__))
if current_dir not in sys.path:
sys.path.insert(0, current_dir)
import amdsmi_rocm_smi_compat
showAllConcise = amdsmi_rocm_smi_compat.showAllConcise
listDevices = amdsmi_rocm_smi_compat.listDevices
initializeRsmi = amdsmi_rocm_smi_compat.initializeRsmi
check_runtime_status = amdsmi_rocm_smi_compat.check_runtime_status
# Initialize AMD SMI
if not initializeRsmi():
logging.error("Failed to initialize AMD SMI")
return
try:
# Get processor handles
deviceList = listDevices()
if not deviceList:
logging.error("No AMD GPU devices found")
return
# Check runtime status (low power state warning)
if not check_runtime_status():
print("\nWARNING: AMD GPU device(s) is/are in a low-power state. Check power control/runtime_status\n")
# Display ROCm-SMI compatible output
showAllConcise(deviceList)
finally:
# Shutdown AMD SMI
try:
amdsmi_interface.amdsmi_shut_down()
except:
pass
except ImportError as e:
logging.error(f"Could not import ROCm-SMI compatibility module: {e}")
logging.error("Make sure amdsmi_rocm_smi_compat.py is in the amdsmi_cli directory")
print("ERROR: ROCm-SMI compatibility mode not available")
except Exception as e:
logging.error(f"Error in ROCm-SMI compatibility mode: {e}")
print(f"ERROR: {e}")
Fai riferimento in un nuovo problema Visualizza Git Blame Copia Permalink