AI/rocm-systems
2
0
フォーク 0
コード イシュー プルリクエスト Actions パッケージ プロジェクト リリース Wiki アクティビティ

[SWDEV-503520] Add amdsmi_get_rocm_version() in python library (#76)

ソースを参照 
Changed amdsmi_get_rocm_version() to be an API in the python library only. 
Updated usage and version detection
Updated path detection of librocm-core.so
Updated docs to reflect both amdsmi_get_rocm_version and amdsmi_get_lib_version() do not require initialization.

Signed-off-by: Arif, Maisam <Maisam.Arif@amd.com>
このコミットが含まれているのは:
Arif, Maisam
2025-02-26 05:45:58 -06:00
committed by GitHub
親 83c88f26d1
コミット 52b3ee2dc6
9個のファイルの変更、191行の追加、120行の削除
Patchファイルをダウンロード Diffファイルをダウンロード すべてのファイルを展開 すべてのファイルを折り畳む
amdsmi_cli/CMakeLists.txt
-3
ファイルの表示
@@ -22,7 +22,6 @@ add_custom_command(
${PY_PACKAGE_DIR}/amdsmi_logger.py
${PY_PACKAGE_DIR}/amdsmi_parser.py
${PY_PACKAGE_DIR}/amdsmi_cli_exceptions.py
${PY_PACKAGE_DIR}/rocm_version.py
${PY_PACKAGE_DIR}/BDF.py
${PY_PACKAGE_DIR}/README.md
${PY_PACKAGE_DIR}/Release_Notes.md
@@ -36,7 +35,6 @@ add_custom_command(
COMMAND ln -Pf ${CMAKE_CURRENT_SOURCE_DIR}/amdsmi_logger.py ${PY_PACKAGE_DIR}/
COMMAND ln -Pf ${CMAKE_CURRENT_SOURCE_DIR}/amdsmi_parser.py ${PY_PACKAGE_DIR}/
COMMAND ln -Pf ${CMAKE_CURRENT_SOURCE_DIR}/amdsmi_cli_exceptions.py ${PY_PACKAGE_DIR}/
COMMAND ln -Pf ${CMAKE_CURRENT_SOURCE_DIR}/rocm_version.py ${PY_PACKAGE_DIR}/
COMMAND ln -Pf ${CMAKE_CURRENT_SOURCE_DIR}/BDF.py ${PY_PACKAGE_DIR}/
COMMAND ln -Pf ${CMAKE_CURRENT_SOURCE_DIR}/README.md ${PY_PACKAGE_DIR}/
COMMAND ln -Pf ${CMAKE_CURRENT_SOURCE_DIR}/Release_Notes.md ${PY_PACKAGE_DIR}/)
@@ -54,7 +52,6 @@ add_custom_target(
${PY_PACKAGE_DIR}/amdsmi_logger.py
${PY_PACKAGE_DIR}/amdsmi_parser.py
${PY_PACKAGE_DIR}/amdsmi_cli_exceptions.py
${PY_PACKAGE_DIR}/rocm_version.py
${PY_PACKAGE_DIR}/BDF.py
${PY_PACKAGE_DIR}/README.md
${PY_PACKAGE_DIR}/Release_Notes.md)
amdsmi_cli/amdsmi_commands.py
+18 -30
ファイルの表示
@@ -19,6 +19,7 @@
# 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 logging
import sys
import threading
@@ -32,7 +33,6 @@ from _version import __version__
from amdsmi_helpers import AMDSMIHelpers
from amdsmi_logger import AMDSMILogger
from amdsmi_cli_exceptions import AmdSmiRequiredCommandException, AmdSmiInvalidParameterException
from rocm_version import get_rocm_version
from amdsmi import amdsmi_interface
from amdsmi import amdsmi_exception
@@ -95,36 +95,14 @@ class AMDSMICommands():
exit_flag = True
if exit_flag:
try:
amdsmi_lib_version = amdsmi_interface.amdsmi_get_lib_version()
amdsmi_lib_version_str = f"{amdsmi_lib_version['year']}.{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()
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'{get_rocm_version()}'
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: {get_rocm_version()}"
# 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()
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:
@@ -136,17 +114,23 @@ class AMDSMICommands():
if cpu_version:
args.cpu_version = cpu_version
# if no args are given, display everything
if not args.gpu_version and not args.cpu_version:
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['year']}.{amdsmi_lib_version['major']}.{amdsmi_lib_version['minor']}.{amdsmi_lib_version['release']}"
rocm_version_str = get_rocm_version()
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}'
@@ -984,7 +968,7 @@ class AMDSMICommands():
static_dict['clock'] = clk_dict
else:
raise amdsmi_exception.AmdSmiParameterException(args.clock, list[str])
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
@@ -4218,6 +4202,7 @@ class AMDSMICommands():
self.logger.store_output(args.gpu, 'perfdeterminism', f"Successfully enabled performance determinism and set GFX clock frequency to {args.perf_determinism}")
if args.compute_partition:
attempted_to_set = "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)))
@@ -4474,6 +4459,9 @@ class AMDSMICommands():
amdsmi_clk_type = amdsmi_interface.AmdSmiClkType.GFX
elif clk_type == "mclk":
amdsmi_clk_type = amdsmi_interface.AmdSmiClkType.MEM
else:
raise ValueError(f"Invalid clock type {clk_type} for {gpu_string}")
clk_tuple = amdsmi_interface.amdsmi_get_clock_info(args.gpu, amdsmi_clk_type)
if lim_type == "min":
amdsmi_cli/amdsmi_helpers.py
+13 -2
ファイルの表示
@@ -29,9 +29,8 @@ import re
import multiprocessing
import json
from typing import List, Union
from enum import Enum
from typing import Set
from typing import List, Set, Union
from amdsmi_init import *
from BDF import BDF
@@ -98,9 +97,11 @@ class AMDSMIHelpers():
def increment_set_count(self):
self._count_of_sets_called += 1
def get_set_count(self):
return self._count_of_sets_called
def is_virtual_os(self):
return self._is_virtual_os
@@ -173,6 +174,16 @@ class AMDSMIHelpers():
return AMDSMI_INIT_FLAG & amdsmi_interface.amdsmi_wrapper.AMDSMI_INIT_AMD_CPUS
def get_rocm_version(self):
try:
rocm_lib_status, rocm_version = amdsmi_interface.amdsmi_get_rocm_version()
if rocm_lib_status is not True:
return "N/A"
return rocm_version
except amdsmi_interface.AmdSmiLibraryException as e:
return "N/A"
def get_cpu_choices(self):
"""Return dictionary of possible CPU choices and string of the output:
Dictionary will be in format: cpus[ID]: Device Handle)
amdsmi_cli/amdsmi_parser.py
+1 -2
ファイルの表示
@@ -33,7 +33,6 @@ from pathlib import Path
from _version import __version__
from amdsmi_helpers import AMDSMIHelpers
from rocm_version import get_rocm_version
import amdsmi_cli_exceptions
@@ -95,7 +94,7 @@ class AMDSMIParser(argparse.ArgumentParser):
version_string = f"Version: {__version__}"
platform_string = f"Platform: {self.helpers.os_info()}"
rocm_version = get_rocm_version()
rocm_version = self.helpers.get_rocm_version()
rocm_version_string = f"ROCm version: {rocm_version}"
program_name = 'amd-smi'
amdsmi_cli/rocm_version.py
-29
ファイルの表示
@@ -1,29 +0,0 @@
import os
import ctypes
from pathlib import Path
# Get the ROCm version for rocm-core library
def get_rocm_version():
try:
librocm_core_file = Path(__file__).resolve().parent.parent.parent / "lib" / "librocm-core.so"
if not librocm_core_file.is_file():
return "N/A"
# python binding
librocm_core = ctypes.CDLL(librocm_core_file)
VerErrors = ctypes.c_uint32
get_rocm_core_version = librocm_core.getROCmVersion
get_rocm_core_version.restype = VerErrors
get_rocm_core_version.argtypes = [ctypes.POINTER(ctypes.c_uint32), ctypes.POINTER(ctypes.c_uint32),ctypes.POINTER(ctypes.c_uint32)]
# call the function
major = ctypes.c_uint32()
minor = ctypes.c_uint32()
patch = ctypes.c_uint32()
if get_rocm_core_version(ctypes.byref(major), ctypes.byref(minor),ctypes.byref(patch)) == 0:
return "%d.%d.%d" % (major.value, minor.value, patch.value)
return "N/A"
except:
return "N/A"