AI/rocm-systems
2
0
Форкнуть 0
Код Задачи Запросы на слияние Действия Пакеты Проекты Релизы Вики Активность

Merge amd-dev into amd-master 20231005

Просмотр исходного кода 
Change-Id: Ifa10a154ea9c0c656208f112c687e75fb63bd6d0
Signed-off-by: Galantsev, Dmitrii <dmitrii.galantsev@amd.com>
Этот коммит содержится в:
Galantsev, Dmitrii
2023-10-05 16:31:11 -05:00
родитель 788bad5f43 656f12e0f3
Коммит 593b6b6513
31 изменённых файлов: 424 добавлений и 151 удалений
Скачать Patch файл Скачать Diff файл Показать все файлы Свернуть все файлы
CMakeLists.txt
+4 -2
Просмотреть файл
@@ -1,7 +1,7 @@
#
# Minimum version of cmake required
#
cmake_minimum_required(VERSION 3.11)
cmake_minimum_required(VERSION 3.14)
set(AMD_SMI "amd_smi")
set(AMD_SMI_COMPONENT "lib${AMD_SMI}")
@@ -49,6 +49,8 @@ set(AMD_SMI_PACKAGE
CACHE STRING "")
project(${AMD_SMI_LIBS_TARGET})
set(CMAKE_CXX_STANDARD 17)
set(CMAKE_CXX_STANDARD_REQUIRED ON)
include(GNUInstallDirs)
@@ -78,7 +80,7 @@ set(CPACK_PACKAGE_DESCRIPTION_SUMMARY
generic_package()
## Compiler flags
set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -Wall -Wextra -fno-rtti -m64 -msse -msse2 -std=c++11")
set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -Wall -Wextra -fno-rtti -m64 -msse -msse2")
# Security options
set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -Wconversion -Wcast-align")
set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -Wformat=2 -fno-common -Wstrict-overflow")
amdsmi_cli/amdsmi_cli.py
+9 -1
Просмотреть файл
@@ -42,6 +42,12 @@ def _print_error(e, destination):
if __name__ == "__main__":
# Disable traceback before possible init errors in AMDSMICommands and AMDSMIParser
if "DEBUG" in sys.argv:
sys.tracebacklimit = 10
else:
sys.tracebacklimit = -1
amd_smi_commands = AMDSMICommands()
amd_smi_parser = AMDSMIParser(amd_smi_commands.version,
amd_smi_commands.list,
@@ -80,7 +86,9 @@ if __name__ == "__main__":
logging.basicConfig(format='%(levelname)s: %(message)s', level=logging_dict[args.loglevel])
# Disable traceback for non-debug log levels
if args.loglevel != "DEBUG":
if args.loglevel == "DEBUG":
sys.tracebacklimit = 10
else:
sys.tracebacklimit = -1
# Execute subcommands
amdsmi_cli/amdsmi_commands.py
+32 -29
Просмотреть файл
@@ -21,6 +21,7 @@
#
import logging
import sys
import threading
import time
@@ -42,7 +43,12 @@ class AMDSMICommands():
try:
self.device_handles = amdsmi_interface.amdsmi_get_processor_handles()
except amdsmi_exception.AmdSmiLibraryException as e:
raise 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)')
sys.exit(-1)
else:
raise e
self.stop = ''
self.all_arguments = False
@@ -220,27 +226,21 @@ class AMDSMICommands():
bus_info['max_pcie_speed'] = pcie_speed_GTs_value
try:
slot_type = amdsmi_interface.amdsmi_topo_get_link_type(args.gpu, args.gpu)['type']
except amdsmi_exception.AmdSmiLibraryException as e:
slot_type = e.get_error_info()
slot_type = bus_info.pop('pcie_slot_type')
if isinstance(slot_type, int):
slot_types = amdsmi_interface.amdsmi_wrapper.amdsmi_pcie_slot_type_t__enumvalues
if slot_type in slot_types:
bus_info['slot_type'] = slot_types[slot_type].replace("AMDSMI_SLOT_TYPE__", "")
else:
bus_info['slot_type'] = "Unknown"
else:
bus_info['slot_type'] = "N/A"
if self.logger.is_human_readable_format():
unit ='GT/s'
bus_info['max_pcie_speed'] = f"{bus_info['max_pcie_speed']} {unit}"
if bus_info['pcie_interface_version'] > 0:
bus_info['pcie_interface_version'] = f"Gen {bus_info['pcie_interface_version']}"
bus_info['slot_type'] = 'XXXX'
if isinstance(slot_type, int):
if slot_type == amdsmi_interface.amdsmi_wrapper.AMDSMI_IOLINK_TYPE_UNDEFINED:
bus_info['slot_type'] = "UNKNOWN"
elif slot_type == amdsmi_interface.amdsmi_wrapper.AMDSMI_IOLINK_TYPE_PCIEXPRESS:
bus_info['slot_type'] = "PCIE"
elif slot_type == amdsmi_interface.amdsmi_wrapper.AMDSMI_IOLINK_TYPE_XGMI:
bus_info['slot_type'] = "XGMI"
except amdsmi_exception.AmdSmiLibraryException as e:
bus_info = "N/A"
logging.debug("Failed to get bus info for gpu %s | %s", gpu_id, e.get_error_info())
@@ -263,18 +263,19 @@ class AMDSMICommands():
if self.helpers.is_linux() and self.helpers.is_baremetal():
if args.board:
static_dict['board'] = {"model_number": "N/A",
"product_serial": "N/A",
"fru_id": "N/A",
"manufacturer_name": "N/A",
"product_name": "N/A"}
try:
board_info = amdsmi_interface.amdsmi_get_gpu_board_info(args.gpu)
board_info['serial_number'] = hex(board_info['serial_number'])
board_info['model_number'] = board_info['model_number'].strip()
board_info['product_name'] = board_info['product_name'].strip()
board_info['manufacturer_name'] = board_info['manufacturer_name'].strip()
board_info.pop('product_serial')
board_info.pop('manufacturer_name')
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:
static_dict['board'] = "N/A"
logging.debug("Failed to get board info for gpu %s | %s", gpu_id, e.get_error_info())
if args.limit:
# Power limits
@@ -390,14 +391,16 @@ class AMDSMICommands():
static_dict['limit'] = limit_info
if args.driver:
driver_info = {"driver_name" : "N/A",
"driver_version" : "N/A",
"driver_date" : "N/A"}
try:
driver_info = {}
driver_info = amdsmi_interface.amdsmi_get_gpu_driver_info(args.gpu)
static_dict['driver'] = driver_info
except amdsmi_exception.AmdSmiLibraryException as e:
static_dict['driver'] = "N/A"
logging.debug("Failed to get driver info for gpu %s | %s", gpu_id, e.get_error_info())
static_dict['driver'] = driver_info
if self.helpers.is_hypervisor() or self.helpers.is_baremetal():
if args.ras:
try:
@@ -1006,7 +1009,7 @@ class AMDSMICommands():
values_dict['ecc_block'] = "N/A"
logging.debug("Failed to get ecc block features for gpu %s | %s", gpu_id, e.get_error_info())
if args.pcie:
pcie_dict = {'current_width': "N/A",
pcie_dict = {'current_lanes': "N/A",
'current_speed': "N/A",
'replay_count' : "N/A",
'current_bandwith_sent': "N/A",
@@ -1022,7 +1025,7 @@ class AMDSMICommands():
pcie_speed_GTs_value = round(pcie_link_status['pcie_speed'] / 1000)
pcie_dict['current_speed'] = pcie_speed_GTs_value
pcie_dict['current_width'] = pcie_link_status['pcie_lanes']
pcie_dict['current_lanes'] = pcie_link_status['pcie_lanes']
if self.logger.is_human_readable_format():
unit = 'GT/s'
amdsmi_cli/amdsmi_helpers.py
+11 -2
Просмотреть файл
@@ -141,8 +141,17 @@ class AMDSMIHelpers():
gpu_choices = {}
gpu_choices_str = ""
# amdsmi_get_processor_handles returns the device_handles storted for gpu_id
device_handles = amdsmi_interface.amdsmi_get_processor_handles()
try:
# amdsmi_get_processor_handles returns the device_handles storted for gpu_id
device_handles = amdsmi_interface.amdsmi_get_processor_handles()
except amdsmi_interface.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 device choices, driver not initialized (amdgpu not found in modules)')
sys.exit(-1)
else:
raise e
for gpu_id, device_handle in enumerate(device_handles):
bdf = amdsmi_interface.amdsmi_get_gpu_device_bdf(device_handle)
uuid = amdsmi_interface.amdsmi_get_gpu_device_uuid(device_handle)
amdsmi_cli/amdsmi_init.py
+18 -14
Просмотреть файл
@@ -36,9 +36,9 @@ from amdsmi import amdsmi_interface
from amdsmi import amdsmi_exception
# Using basic python logging for user errors and development
logging.basicConfig(format='%(levelname)s: %(message)s', level=logging.ERROR) # User level logging
logging.basicConfig(format="%(levelname)s: %(message)s", level=logging.ERROR) # User level logging
# This traceback limit only affects this file, once the code hit's the cli portion it get's reset to the user's preference
sys.tracebacklimit = -1 # Disable traceback for user errors
sys.tracebacklimit = -1 # Disable traceback when raising errors
# On initial import set initialized variable
AMDSMI_INITIALIZED = False
@@ -47,11 +47,9 @@ AMD_VENDOR_ID = 4098
def check_amdgpu_driver():
""" Returns true if amdgpu is found in the list of initialized modules """
amd_gpu_status_file = Path("/sys/module/amdgpu/initstate")
if amd_gpu_status_file.exists():
if amd_gpu_status_file.read_text(encoding='ascii').strip() == 'live':
if amd_gpu_status_file.read_text(encoding="ascii").strip() == "live":
return True
return False
@@ -61,17 +59,22 @@ def init_amdsmi(flag=amdsmi_interface.AmdSmiInitFlags.INIT_AMD_GPUS):
Raises:
err: AmdSmiLibraryException if not successful
"""
# Check if amdgpu driver is up & Handle error gracefully
# # Check if amdgpu driver is up & Handle error gracefully
if check_amdgpu_driver():
# Only init AMD GPUs for now, waiting for future support for AMD CPUs
try:
amdsmi_interface.amdsmi_init(flag)
except (amdsmi_interface.AmdSmiLibraryException, amdsmi_interface.AmdSmiParameterException) as err:
raise err
logging.debug('AMDSMI initialized successfully')
except (amdsmi_interface.AmdSmiLibraryException, amdsmi_interface.AmdSmiParameterException) 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("Driver not loaded (amdgpu not found in modules)")
sys.exit(-1)
else:
raise e
logging.debug("AMDSMI initialized successfully, but initstate was not live")
else:
logging.error('Driver not initialized (amdgpu not found in modules)')
exit(-1)
logging.error("Driver not found (amdgpu not found in modules)")
sys.exit(-1)
def shut_down_amdsmi():
@@ -82,12 +85,13 @@ def shut_down_amdsmi():
"""
try:
amdsmi_interface.amdsmi_shut_down()
except amdsmi_exception.AmdSmiLibraryException as err:
raise err
except amdsmi_exception.AmdSmiLibraryException as e:
logging.error("Unable to cleanly shut down amd-smi-lib")
raise e
def signal_handler(sig, frame):
logging.debug(f'Handling signal: {sig}')
logging.debug(f"Handling signal: {sig}")
sys.exit(0)
amdsmi_cli/amdsmi_parser.py
+1 -1
Просмотреть файл
@@ -348,7 +348,7 @@ class AMDSMIParser(argparse.ArgumentParser):
err_records_help = "All error records information"
# Create firmware subparser
firmware_parser = subparsers.add_parser('firmware', help=firmware_help, description=firmware_subcommand_help)
firmware_parser = subparsers.add_parser('firmware', help=firmware_help, description=firmware_subcommand_help, aliases=['ucode'])
firmware_parser._optionals.title = firmware_optionals_title
firmware_parser.formatter_class=lambda prog: argparse.RawTextHelpFormatter(prog, max_help_position=80, width=90)
firmware_parser.set_defaults(func=func)
cmake_modules/utils.cmake
+1 -1
Просмотреть файл
@@ -52,7 +52,7 @@ function( parse_version VERSION_STRING )
string ( SUBSTRING ${VERSION_STRING} ${STRING_INDEX} -1 VERSION_BUILD )
endif ()
string ( REGEX MATCHALL "[0123456789]+" VERSIONS ${VERSION_STRING} )
string ( REGEX MATCHALL "[0-9]+" VERSIONS ${VERSION_STRING} )
list ( LENGTH VERSIONS VERSION_COUNT )
if ( ${VERSION_COUNT} GREATER 0)
example/CMakeLists.txt.in
+3 -5
Просмотреть файл
@@ -1,13 +1,11 @@
cmake_minimum_required(VERSION 3.11)
cmake_minimum_required(VERSION 3.14)
option(CMAKE_VERBOSE_MAKEFILE "Enable verbose output" ON)
option(CMAKE_EXPORT_COMPILE_COMMANDS "Export compile commands for linters and autocompleters" ON)
project(main LANGUAGES CXX)
set(CMAKE_CXX_STANDARD
11
CACHE STRING "The C++ standard to use")
set(CMAKE_CXX_STANDARD 17)
set(CMAKE_CXX_STANDARD_REQUIRED ON)
find_package(amd_smi
HINTS
example/amd_smi_drm_example.cc
+6 -4
Просмотреть файл
@@ -322,6 +322,7 @@ int main() {
ret = amdsmi_get_gpu_driver_info(processor_handles[j], &driver_info);
CHK_AMDSMI_RET(ret)
printf(" Output of amdsmi_get_gpu_driver_info:\n");
printf("\tDriver name: %s\n", driver_info.driver_name);
printf("\tDriver version: %s\n", driver_info.driver_version);
printf("\tDriver date: %s\n\n", driver_info.driver_date);
@@ -379,9 +380,10 @@ int main() {
ret = amdsmi_get_pcie_link_status(processor_handles[j], &pcie_info);
CHK_AMDSMI_RET(ret)
printf(" Output of amdsmi_get_pcie_link_status:\n");
printf("\tPCIe lanes: %d\n", pcie_info.pcie_lanes);
printf("\tPCIe speed: %d\n\n", pcie_info.pcie_speed);
printf("\tPCIe Interface Version: %d\n\n", pcie_info.pcie_interface_version);
printf("\tCurrent PCIe lanes: %d\n", pcie_info.pcie_lanes);
printf("\tCurrent PCIe speed: %d\n", pcie_info.pcie_speed);
printf("\tCurrent PCIe Interface Version: %d\n", pcie_info.pcie_interface_version);
printf("\tPCIe slot type: %d\n\n", pcie_info.pcie_slot_type);
// Get PCIe caps
amdsmi_pcie_info_t pcie_caps_info = {};
@@ -389,7 +391,7 @@ int main() {
CHK_AMDSMI_RET(ret)
printf(" Output of amdsmi_get_pcie_link_caps:\n");
printf("\tPCIe max lanes: %d\n", pcie_caps_info.pcie_lanes);
printf("\tPCIe max speed: %d\n\n", pcie_caps_info.pcie_speed);
printf("\tPCIe max speed: %d\n", pcie_caps_info.pcie_speed);
printf("\tPCIe Interface Version: %d\n\n", pcie_caps_info.pcie_interface_version);
// Get VRAM temperature limit
include/amd_smi/amdsmi.h
+14 -1
Просмотреть файл
@@ -444,6 +444,7 @@ typedef struct{
typedef struct {
char driver_name[AMDSMI_MAX_STRING_LENGTH];
char driver_version[AMDSMI_MAX_STRING_LENGTH];
char driver_date[AMDSMI_MAX_STRING_LENGTH];
} amdsmi_driver_info_t;
@@ -1150,6 +1151,16 @@ typedef struct {
uint64_t reserved[2];
} amdsmi_error_count_t;
/**
* @brief This is a enum translation for pcie_slot_type
*/
typedef enum {
AMDSMI_SLOT_TYPE__PCIE = 0,
AMDSMI_SLOT_TYPE__CEM = 1,
AMDSMI_SLOT_TYPE__OAM = 2,
AMDSMI_SLOT_TYPE__RESERVED = 3,
} amdsmi_pcie_slot_type_t;
/**
* @brief This structure holds pcie info.
*/
@@ -1157,8 +1168,10 @@ typedef struct {
uint16_t pcie_lanes;
uint32_t pcie_speed;
uint32_t pcie_interface_version;
uint32_t reserved[5];
amdsmi_pcie_slot_type_t pcie_slot_type; // 0: PCIE, 1: CEM, 2: OAM, 3: Reserved
uint32_t reserved[4];
} amdsmi_pcie_info_t;
/**
* @brief This structure contains information specific to a process.
*/
include/amd_smi/impl/amd_smi_drm.h
+1
Просмотреть файл
@@ -75,6 +75,7 @@ class AMDSmiDrm {
amdsmi_status_t amdgpu_query_hw_ip(int fd, unsigned info_id,
unsigned hw_ip_type, unsigned size, void *value);
amdsmi_status_t amdgpu_query_vbios(int fd, void *info);
amdsmi_status_t amdgpu_query_driver_name(int fd, std::string& driver_name);
amdsmi_status_t amdgpu_query_driver_date(int fd, std::string& driver_date);
private:
include/amd_smi/impl/amd_smi_gpu_device.h
+1
Просмотреть файл
@@ -81,6 +81,7 @@ class AMDSmiGPUDevice: public AMDSmiProcessor {
amdsmi_status_t amdgpu_query_fw(unsigned info_id, unsigned fw_type,
unsigned size, void *value) const;
amdsmi_status_t amdgpu_query_vbios(void *info) const;
amdsmi_status_t amdgpu_query_driver_name(std::string& name) const;
amdsmi_status_t amdgpu_query_driver_date(std::string& date) const;
private:
uint32_t gpu_id_;
py-interface/README.md
+7 -1
Просмотреть файл
@@ -310,7 +310,13 @@ Input parameters:
* `processor_handle` dev for which to query
Output: Driver info that is handling the device
Output: Dictionary with fields
Field | Content
---|---
`driver_name` | driver name
`driver_version` | driver_version
`driver_date` | driver_date
Exceptions that can be thrown by `amdsmi_get_gpu_driver_info` function:
py-interface/amdsmi_exception.py
+1 -1
Просмотреть файл
@@ -72,7 +72,7 @@ class AmdSmiLibraryException(AmdSmiException):
amdsmi_wrapper.AMDSMI_STATUS_NOT_FOUND : "AMDSMI_STATUS_NOT_FOUND - Device Not found",
amdsmi_wrapper.AMDSMI_STATUS_NOT_INIT : "AMDSMI_STATUS_NOT_INIT - Device not initialized",
amdsmi_wrapper.AMDSMI_STATUS_NO_SLOT : "AMDSMI_STATUS_NO_SLOT - No more free slot",
amdsmi_wrapper.AMDSMI_STATUS_DRIVER_NOT_LOADED : "AMDSMI_STATUS_DRIVER_NOT_LOADED - Processor driver not loaded",
amdsmi_wrapper.AMDSMI_STATUS_DRIVER_NOT_LOADED : "AMDSMI_STATUS_DRIVER_NOT_LOADED - Driver not loaded",
amdsmi_wrapper.AMDSMI_STATUS_NO_DATA : "AMDSMI_STATUS_NO_DATA - No data was found for given input",
amdsmi_wrapper.AMDSMI_STATUS_INSUFFICIENT_SIZE : "AMDSMI_STATUS_INSUFFICIENT_SIZE - Insufficient size for operation",
amdsmi_wrapper.AMDSMI_STATUS_UNEXPECTED_SIZE : "AMDSMI_STATUS_UNEXPECTED_SIZE - unexpected size of data was read",
py-interface/amdsmi_interface.py
+10 -7
Просмотреть файл
@@ -823,11 +823,11 @@ def amdsmi_get_gpu_board_info(
)
return {
"serial_number": board_info.serial_number,
"model_number": board_info.model_number.decode("utf-8"),
"product_serial": board_info.product_serial.decode("utf-8"),
"product_name": board_info.product_name.decode("utf-8"),
"manufacturer_name" : board_info.product_name.decode("utf-8")
"model_number": board_info.model_number.decode("utf-8").strip(),
"product_serial": board_info.serial_number,
"fru_id": board_info.fru_id.decode("utf-8").strip(),
"manufacturer_name" : board_info.manufacturer_name.decode("utf-8").strip(),
"product_name": board_info.product_name.decode("utf-8").strip()
}
@@ -962,6 +962,7 @@ def amdsmi_get_gpu_driver_info(
)
return {
"driver_name": info.driver_name.decode("utf-8"),
"driver_version": info.driver_version.decode("utf-8"),
"driver_date": info.driver_date.decode("utf-8")
}
@@ -1061,7 +1062,8 @@ def amdsmi_get_pcie_link_status(
return {"pcie_speed": pcie_info.pcie_speed,
"pcie_lanes": pcie_info.pcie_lanes,
"pcie_interface_version": pcie_info.pcie_interface_version}
"pcie_interface_version": pcie_info.pcie_interface_version,
"pcie_slot_type": pcie_info.pcie_slot_type}
def amdsmi_get_pcie_link_caps(
processor_handle: amdsmi_wrapper.amdsmi_processor_handle,
@@ -1079,7 +1081,8 @@ def amdsmi_get_pcie_link_caps(
return {"max_pcie_speed": pcie_info.pcie_speed,
"max_pcie_lanes": pcie_info.pcie_lanes,
"pcie_interface_version": pcie_info.pcie_interface_version}
"pcie_interface_version": pcie_info.pcie_interface_version,
"pcie_slot_type": pcie_info.pcie_slot_type}
def amdsmi_get_processor_handle_from_bdf(bdf):
py-interface/amdsmi_wrapper.py
+21 -4
Просмотреть файл
@@ -774,6 +774,7 @@ class struct_amdsmi_driver_info_t(Structure):
struct_amdsmi_driver_info_t._pack_ = 1 # source:False
struct_amdsmi_driver_info_t._fields_ = [
('driver_name', ctypes.c_char * 64),
('driver_version', ctypes.c_char * 64),
('driver_date', ctypes.c_char * 64),
]
@@ -1424,6 +1425,19 @@ struct_amdsmi_error_count_t._fields_ = [
]
amdsmi_error_count_t = struct_amdsmi_error_count_t
# values for enumeration 'amdsmi_pcie_slot_type_t'
amdsmi_pcie_slot_type_t__enumvalues = {
0: 'AMDSMI_SLOT_TYPE__PCIE',
1: 'AMDSMI_SLOT_TYPE__CEM',
2: 'AMDSMI_SLOT_TYPE__OAM',
3: 'AMDSMI_SLOT_TYPE__RESERVED',
}
AMDSMI_SLOT_TYPE__PCIE = 0
AMDSMI_SLOT_TYPE__CEM = 1
AMDSMI_SLOT_TYPE__OAM = 2
AMDSMI_SLOT_TYPE__RESERVED = 3
amdsmi_pcie_slot_type_t = ctypes.c_uint32 # enum
class struct_amdsmi_pcie_info_t(Structure):
pass
@@ -1433,7 +1447,8 @@ struct_amdsmi_pcie_info_t._fields_ = [
('PADDING_0', ctypes.c_ubyte * 2),
('pcie_speed', ctypes.c_uint32),
('pcie_interface_version', ctypes.c_uint32),
('reserved', ctypes.c_uint32 * 5),
('pcie_slot_type', amdsmi_pcie_slot_type_t),
('reserved', ctypes.c_uint32 * 4),
]
amdsmi_pcie_info_t = struct_amdsmi_pcie_info_t
@@ -1804,7 +1819,9 @@ __all__ = \
'AMDSMI_RAS_ERR_STATE_INVALID', 'AMDSMI_RAS_ERR_STATE_LAST',
'AMDSMI_RAS_ERR_STATE_MULT_UC', 'AMDSMI_RAS_ERR_STATE_NONE',
'AMDSMI_RAS_ERR_STATE_PARITY', 'AMDSMI_RAS_ERR_STATE_POISON',
'AMDSMI_RAS_ERR_STATE_SING_C', 'AMDSMI_STATUS_ADDRESS_FAULT',
'AMDSMI_RAS_ERR_STATE_SING_C', 'AMDSMI_SLOT_TYPE__CEM',
'AMDSMI_SLOT_TYPE__OAM', 'AMDSMI_SLOT_TYPE__PCIE',
'AMDSMI_SLOT_TYPE__RESERVED', 'AMDSMI_STATUS_ADDRESS_FAULT',
'AMDSMI_STATUS_API_FAILED', 'AMDSMI_STATUS_BUSY',
'AMDSMI_STATUS_DRIVER_NOT_LOADED', 'AMDSMI_STATUS_DRM_ERROR',
'AMDSMI_STATUS_FAIL_LOAD_MODULE',
@@ -1954,8 +1971,8 @@ __all__ = \
'amdsmi_mm_ip_t', 'amdsmi_od_vddc_point_t',
'amdsmi_od_volt_curve_t', 'amdsmi_od_volt_freq_data_t',
'amdsmi_pcie_bandwidth_t', 'amdsmi_pcie_info_t',
'amdsmi_power_cap_info_t', 'amdsmi_power_info_t',
'amdsmi_power_profile_preset_masks_t',
'amdsmi_pcie_slot_type_t', 'amdsmi_power_cap_info_t',
'amdsmi_power_info_t', 'amdsmi_power_profile_preset_masks_t',
'amdsmi_power_profile_status_t', 'amdsmi_proc_info_t',
'amdsmi_process_handle_t', 'amdsmi_process_info_t',
'amdsmi_processor_handle', 'amdsmi_range_t',
rocm_smi/include/rocm_smi/rocm_smi.h
+33
Просмотреть файл
@@ -625,6 +625,19 @@ typedef enum {
typedef rsmi_freq_ind_t rsmi_freq_ind;
/// \endcond
/**
* @brief The values of this enum are used as PCIe slot type.
*/
typedef enum {
RSMI_PCIE_SLOT_PCIE = 0,
RSMI_PCIE_SLOT_CEM = 1,
RSMI_PCIE_SLOT_OAM = 2,
RSMI_PCIE_SLOT_UNKNOWN = 3 //!< An unknown
} rsmi_pcie_slot_type_t;
/// \cond Ignore in docs.
typedef rsmi_pcie_slot_type_t rsmi_pcie_slot_type;
/// \endcond
/**
* @brief The values of this enum are used to identify the various firmware
@@ -1391,6 +1404,26 @@ rsmi_status_t rsmi_dev_vendor_name_get(uint32_t dv_ind, char *name,
rsmi_status_t rsmi_dev_vram_vendor_get(uint32_t dv_ind, char *brand,
uint32_t len);
/**
* @brief Get the PCIe slot type of a gpu device.
*
* @details Given a device index @p dv_ind, a pointer to a caller provided
* char buffer @p type, this function will write the PCIe slot type of the
* device to @p type.
*
*
* @param[in] dv_ind a device index
*
* @param[inout] type a pointer to a caller provided buffer to which the
* type info will be written
*
* @retval ::RSMI_STATUS_SUCCESS is returned upon successful call.
*
*/
rsmi_status_t rsmi_dev_pcie_slot_type_get(uint32_t dv_ind,
rsmi_pcie_slot_type_t* type);
/**
* @brief Get the serial number string for a device
*
rocm_smi/include/rocm_smi/rocm_smi_device.h
+4
Просмотреть файл
@@ -105,6 +105,7 @@ enum DevInfoTypes {
kDevDevRevID,
kDevDevProdName,
kDevDevProdNum,
kDevBoardInfo,
kDevVendorID,
kDevSubSysDevID,
kDevSubSysVendorID,
@@ -200,6 +201,9 @@ class Device {
int readDevInfo(DevInfoTypes type, std::vector<std::string> *retVec);
int readDevInfo(DevInfoTypes type, std::size_t b_size,
void *p_binary_data);
// Get the property from a file which may contain multiple properties.
int readDevInfo(DevInfoTypes type, const std::string& property,
std::string& value);
int writeDevInfo(DevInfoTypes type, uint64_t val);
int writeDevInfo(DevInfoTypes type, std::string val);
rocm_smi/include/rocm_smi/rocm_smi_utils.h
+3
Просмотреть файл
@@ -269,6 +269,9 @@ class ScopedAcquire {
// In VM environment, the /proc/cpuinfo set hypervisor flag by default
bool is_vm_guest();
// trim a string
std::string trim(const std::string &s);
} // namespace smi
} // namespace amd
rocm_smi/python_smi_tools/rocm_smi.py
+38 -5
Просмотреть файл
@@ -30,8 +30,8 @@ from rsmiBindings import *
# Minor version - Increment when adding a new feature, set to 0 when major is incremented
# Patch version - Increment when adding a fix, set to 0 when minor is incremented
SMI_MAJ = 1
SMI_MIN = 4
SMI_PAT = 1
SMI_MIN = 5
SMI_PAT = 0
__version__ = '%s.%s.%s' % (SMI_MAJ, SMI_MIN, SMI_PAT)
# Set to 1 if an error occurs
@@ -1675,6 +1675,32 @@ def setNPSMode(deviceList, npsMode):
printErrLog(device, 'Failed to retrieve NPS mode, even though device supports it.')
printLogSpacer()
def showVersion(isCSV=False):
values = { 'ROCM-SMI version': __version__ }
version = rsmi_version_t()
status = rocmsmi.rsmi_version_get(byref(version))
if status == 0:
version_string = "%u.%u.%u" % (version.major, version.minor, version.patch)
values['ROCM-SMI-LIB version'] = version_string
if isCSV:
print('name, value')
for k in values.keys():
print('%s, %s' % (k, values[k]))
return
if PRINT_JSON:
temp_str = '{\n'
for k in values.keys():
temp_str += ' "%s": "%s",\n' % (k, values[k])
if len(values.keys()) > 1:
# replace ',\n' with '\n}'
temp_str = temp_str[:-2]
temp_str += '\n}'
print(temp_str)
return
for k in values.keys():
print('%s: %s' % (k, values[k]))
def showAllConcise(deviceList):
""" Display critical info for all devices in a concise format
@@ -2052,7 +2078,7 @@ def showFwInfo(deviceList, fwType):
ret = rocmsmi.rsmi_dev_firmware_version_get(device, fw_block_names_l.index(fw_name), byref(fw_ver))
if rsmi_ret_ok(ret, device, 'get_firmware_version_' + str(fw_name)):
# The VCN, VCE, UVD, SOS and ASD firmware's value needs to be in hexadecimal
if fw_name in ['VCN', 'VCE', 'UVD', 'SOS', 'ASD']:
if fw_name in ['VCN', 'VCE', 'UVD', 'SOS', 'ASD', 'MES', 'MES KIQ']:
printLog(device, '%s firmware version' % (fw_name),
'\t0x%s' % (str(hex(fw_ver.value))[2:].zfill(8)))
# The TA XGMI, TA RAS, and SMC firmware's hex value looks like 0x12345678
@@ -2955,7 +2981,7 @@ def showTempGraph(deviceList):
printLogSpacer()
def showVersion(deviceList, component):
def showDriverVersion(deviceList, component):
""" Display the software version for the specified component
@param deviceList: List of DRM devices (can be a single-item list)
@@ -3614,6 +3640,7 @@ if __name__ == '__main__':
parser = argparse.ArgumentParser(
description='AMD ROCm System Management Interface | ROCM-SMI version: %s' % __version__,
formatter_class=lambda prog: argparse.HelpFormatter(prog, max_help_position=90, width=120))
groupVersion = parser.add_argument_group()
groupDev = parser.add_argument_group()
groupDisplayOpt = parser.add_argument_group('Display Options')
groupDisplayTop = parser.add_argument_group('Topology')
@@ -3627,6 +3654,7 @@ if __name__ == '__main__':
groupResponse = parser.add_argument_group('Auto-response options')
groupActionOutput = parser.add_argument_group('Output options')
groupVersion.add_argument('-V', '--version', help='Show version information', action='store_true')
groupDev.add_argument('-d', '--device', help='Execute command on specified device', type=int, nargs='+')
groupDisplayOpt.add_argument('--alldevices', action='store_true') # ------------- function deprecated, no help menu
groupDisplayOpt.add_argument('--showhw', help='Show Hardware details', action='store_true')
@@ -3775,11 +3803,16 @@ if __name__ == '__main__':
# Must set PRINT_JSON early so the prints can be silenced
if args.json or args.csv:
PRINT_JSON = True
# Initialize rsmiBindings
rocmsmi = initRsmiBindings(silent=PRINT_JSON)
# Initialize the rocm SMI library
initializeRsmi()
if args.version:
showVersion(isCSV=args.csv)
sys.exit()
logging.basicConfig(format='%(levelname)s: %(message)s', level=logging.WARNING)
if args.loglevel is not None:
numericLogLevel = getattr(logging, args.loglevel.upper(), logging.WARNING)
@@ -3878,7 +3911,7 @@ if __name__ == '__main__':
if args.showhw:
showAllConciseHw(deviceList)
if args.showdriverversion:
showVersion(deviceList, rsmi_sw_component_t.RSMI_SW_COMP_DRIVER)
showDriverVersion(deviceList, rsmi_sw_component_t.RSMI_SW_COMP_DRIVER)
if args.showtempgraph:
showTempGraph(deviceList)
if args.showid:
rocm_smi/python_smi_tools/rsmiBindings.py
+17 -17
Просмотреть файл
@@ -37,8 +37,6 @@ def initRsmiBindings(silent=False):
print_silent('Using lib from %s' % path_librocm)
else:
print('Unable to find librocm_smi64.so.@VERSION_MAJOR@')
else:
print_silent('Library loaded from: %s ' % path_librocm)
# ----------> TODO: Support static libs as well as SO
try:
@@ -419,25 +417,27 @@ class rsmi_fw_block_t(c_int):
RSMI_FW_BLOCK_ME = 4
RSMI_FW_BLOCK_MEC = 5
RSMI_FW_BLOCK_MEC2 = 6
RSMI_FW_BLOCK_PFP = 7
RSMI_FW_BLOCK_RLC = 8
RSMI_FW_BLOCK_RLC_SRLC = 9
RSMI_FW_BLOCK_RLC_SRLG = 10
RSMI_FW_BLOCK_RLC_SRLS = 11
RSMI_FW_BLOCK_SDMA = 12
RSMI_FW_BLOCK_SDMA2 = 13
RSMI_FW_BLOCK_SMC = 14
RSMI_FW_BLOCK_SOS = 15
RSMI_FW_BLOCK_TA_RAS = 16
RSMI_FW_BLOCK_TA_XGMI = 17
RSMI_FW_BLOCK_UVD = 18
RSMI_FW_BLOCK_VCE = 19
RSMI_FW_BLOCK_VCN = 20
RSMI_FW_BLOCK_MES = 7
RSMI_FW_BLOCK_MES_KIQ = 8
RSMI_FW_BLOCK_PFP = 9
RSMI_FW_BLOCK_RLC = 10
RSMI_FW_BLOCK_RLC_SRLC = 11
RSMI_FW_BLOCK_RLC_SRLG = 12
RSMI_FW_BLOCK_RLC_SRLS = 13
RSMI_FW_BLOCK_SDMA = 14
RSMI_FW_BLOCK_SDMA2 = 15
RSMI_FW_BLOCK_SMC = 16
RSMI_FW_BLOCK_SOS = 17
RSMI_FW_BLOCK_TA_RAS = 18
RSMI_FW_BLOCK_TA_XGMI = 19
RSMI_FW_BLOCK_UVD = 20
RSMI_FW_BLOCK_VCE = 21
RSMI_FW_BLOCK_VCN = 22
RSMI_FW_BLOCK_LAST = RSMI_FW_BLOCK_VCN
# The following list correlated to the rsmi_fw_block_t
fw_block_names_l = ['ASD', 'CE', 'DMCU', 'MC', 'ME', 'MEC', 'MEC2', 'PFP',\
fw_block_names_l = ['ASD', 'CE', 'DMCU', 'MC', 'ME', 'MEC', 'MEC2', 'MES', 'MES KIQ', 'PFP',\
'RLC', 'RLC SRLC', 'RLC SRLG', 'RLC SRLS', 'SDMA', 'SDMA2',\
'SMC', 'SOS', 'TA RAS', 'TA XGMI', 'UVD', 'VCE', 'VCN']
rocm_smi/src/rocm_smi.cc
+23
Просмотреть файл
@@ -910,6 +910,29 @@ rsmi_dev_vendor_id_get(uint32_t dv_ind, uint16_t *id) {
return get_id(dv_ind, amd::smi::kDevVendorID, id);
}
rsmi_status_t
rsmi_dev_pcie_slot_type_get(uint32_t dv_ind, rsmi_pcie_slot_type_t* type) {
TRY
std::ostringstream ss;
ss << __PRETTY_FUNCTION__ << "| ======= start =======";
LOG_TRACE(ss);
CHK_SUPPORT_NAME_ONLY(type)
DEVICE_MUTEX
std::string value;
int ret = dev->readDevInfo(amd::smi::kDevBoardInfo, "type", value);
if (ret != 0) return RSMI_STATUS_NOT_SUPPORTED;
*type = RSMI_PCIE_SLOT_PCIE;
if (value.compare("oam") == 0) *type=RSMI_PCIE_SLOT_OAM;
else if (value.compare("cem") == 0 ) *type=RSMI_PCIE_SLOT_CEM;
else if (value.compare("unknown") == 0 ) *type=RSMI_PCIE_SLOT_UNKNOWN;
return RSMI_STATUS_SUCCESS;
CATCH
}
rsmi_status_t
rsmi_dev_subsystem_vendor_id_get(uint32_t dv_ind, uint16_t *id) {
std::ostringstream ss;
rocm_smi/src/rocm_smi_device.cc
+31
Просмотреть файл
@@ -87,6 +87,7 @@ static const char *kDevDevProdNumFName = "product_number";
static const char *kDevDevIDFName = "device";
static const char *kDevDevRevIDFName = "revision";
static const char *kDevVendorIDFName = "vendor";
static const char *kDevBoardInfoFName = "board_info";
static const char *kDevSubSysDevIDFName = "subsystem_device";
static const char *kDevSubSysVendorIDFName = "subsystem_vendor";
static const char *kDevOverDriveLevelFName = "pp_sclk_od";
@@ -238,6 +239,7 @@ static const std::map<DevInfoTypes, const char *> kDevAttribNameMap = {
{kDevPerfLevel, kDevPerfLevelFName},
{kDevOverDriveLevel, kDevOverDriveLevelFName},
{kDevMemOverDriveLevel, kDevMemOverDriveLevelFName},
{kDevBoardInfo, kDevBoardInfoFName},
{kDevDevProdName, kDevDevProdNameFName},
{kDevDevProdNum, kDevDevProdNumFName},
{kDevDevID, kDevDevIDFName},
@@ -388,6 +390,7 @@ static const std::map<const char *, dev_depends_t> kDevFuncDependsMap = {
{"rsmi_dev_name_get", {{kDevVendorIDFName,
kDevDevIDFName}, {}}},
{"rsmi_dev_sku_get", {{kDevDevProdNumFName}, {}}},
{"rsmi_dev_pcie_slot_type_get", {{kDevBoardInfoFName}, {}}},
{"rsmi_dev_brand_get", {{kDevVendorIDFName,
kDevVBiosVerFName}, {}}},
{"rsmi_dev_vendor_name_get", {{kDevVendorIDFName}, {}}},
@@ -1003,6 +1006,34 @@ int Device::readDevInfo(DevInfoTypes type, uint64_t *val) {
return 0;
}
// Read a property from a file which may contain multiple properties
int Device::readDevInfo(DevInfoTypes type, const std::string& property,
std::string& value) {
std::vector<std::string> val;
int ret = 0;
switch (type) {
case kDevBoardInfo:
ret = readDevInfoMultiLineStr(type, &val);
break;
default:
return EINVAL;
}
if (ret != 0) return ret;
// Find the property from the file
for (unsigned int i = 0; i < val.size(); i++) {
auto pos = val[i].find(":"); // delimiter
if (pos == std::string::npos) continue;
auto name = trim(val[i].substr(0, pos));
if (name != property) continue;
value = trim(val[i].substr(pos+1));
return 0;
}
return EINVAL;
}
int Device::readDevInfo(DevInfoTypes type, std::vector<std::string> *val) {
assert(val != nullptr);
rocm_smi/src/rocm_smi_main.cc
+1
Просмотреть файл
@@ -86,6 +86,7 @@ amd::smi::RocmSMI::devInfoTypesStrings = {
{amd::smi::kDevDevID, amdSMI + "kDevDevID"},
{amd::smi::kDevDevRevID, amdSMI + "kDevDevRevID"},
{amd::smi::kDevDevProdName, amdSMI + "kDevDevProdName"},
{amd::smi::kDevBoardInfo, amdSMI + "kDevBoardInfo"},
{amd::smi::kDevDevProdNum, amdSMI + "kDevDevProdNum"},
{amd::smi::kDevVendorID, amdSMI + "kDevVendorID"},
{amd::smi::kDevSubSysDevID, amdSMI + "kDevSubSysDevID"},
src/amd_smi/amd_smi.cc
+38 -5
Просмотреть файл
@@ -457,7 +457,7 @@ amdsmi_status_t amdsmi_get_gpu_board_info(amdsmi_processor_handle processor_hand
return r;
if (gpu_device->check_if_drm_is_supported()) {
// Get from sys file
// Populate product_serial, product_name, & product_number from sysfs
status = smi_amdgpu_get_board_info(gpu_device, board_info);
}
else {
@@ -472,6 +472,9 @@ amdsmi_status_t amdsmi_get_gpu_board_info(amdsmi_processor_handle processor_hand
}
}
// Get FRU ID
// Get manufacturer name
return AMDSMI_STATUS_SUCCESS;
}
@@ -1718,18 +1721,30 @@ amdsmi_status_t amdsmi_get_gpu_driver_info(amdsmi_processor_handle processor_han
return r;
int length = AMDSMI_MAX_STRING_LENGTH;
// Get the driver version
status = smi_amdgpu_get_driver_version(gpu_device,
&length, info->driver_version);
// Get the driver date
std::string driver_date;
status = gpu_device->amdgpu_query_driver_date(driver_date);
if (status != AMDSMI_STATUS_SUCCESS) return r;
if (status != AMDSMI_STATUS_SUCCESS)
return r;
// Reformat the driver date from 20150101 to 2015/01/01 00:00
if (driver_date.length() == 8) {
driver_date = driver_date.substr(0, 4) + "/" + driver_date.substr(4, 2)
+ "/" + driver_date.substr(6, 2) + " 00:00";
}
strncpy(info->driver_date, driver_date.c_str(), AMDSMI_MAX_STRING_LENGTH-1);
// Get the driver name
std::string driver_name;
status = gpu_device->amdgpu_query_driver_name(driver_name);
if (status != AMDSMI_STATUS_SUCCESS)
return r;
strncpy(info->driver_name, driver_name.c_str(), AMDSMI_MAX_STRING_LENGTH-1);
return status;
}
@@ -1813,7 +1828,16 @@ amdsmi_get_pcie_link_status(amdsmi_processor_handle processor_handle, amdsmi_pci
info->pcie_interface_version = 0;
}
return status;
// default to PCIe
info->pcie_slot_type = AMDSMI_SLOT_TYPE__PCIE;
rsmi_pcie_slot_type_t slot_type;
status = rsmi_wrapper(rsmi_dev_pcie_slot_type_get,
processor_handle, &slot_type);
if (status == AMDSMI_STATUS_SUCCESS) {
info->pcie_slot_type = static_cast<amdsmi_pcie_slot_type_t>(slot_type);
}
return AMDSMI_STATUS_SUCCESS;
}
amdsmi_status_t amdsmi_get_pcie_link_caps(amdsmi_processor_handle processor_handle, amdsmi_pcie_info_t *info) {
@@ -1887,7 +1911,16 @@ amdsmi_status_t amdsmi_get_pcie_link_caps(amdsmi_processor_handle processor_hand
info->pcie_interface_version = 0;
}
return status;
// default to PCIe
info->pcie_slot_type = AMDSMI_SLOT_TYPE__PCIE;
rsmi_pcie_slot_type_t slot_type;
status = rsmi_wrapper(rsmi_dev_pcie_slot_type_get,
processor_handle, &slot_type);
if (status == AMDSMI_STATUS_SUCCESS) {
info->pcie_slot_type = static_cast<amdsmi_pcie_slot_type_t>(slot_type);
}
return AMDSMI_STATUS_SUCCESS;
}
amdsmi_status_t amdsmi_get_processor_handle_from_bdf(amdsmi_bdf_t bdf,
src/amd_smi/amd_smi_drm.cc
+12
Просмотреть файл
@@ -197,6 +197,18 @@ amdsmi_status_t AMDSmiDrm::cleanup() {
return AMDSMI_STATUS_SUCCESS;
}
amdsmi_status_t AMDSmiDrm::amdgpu_query_driver_name(int fd, std::string& driver_name) {
// RAII handler
using drm_version_ptr = std::unique_ptr<drmVersion,
decltype(&drmFreeVersion)>;
std::lock_guard<std::mutex> guard(drm_mutex_);
auto version = drm_version_ptr(
drm_get_version_(fd), drm_free_version_);
if (version == nullptr) return AMDSMI_STATUS_DRM_ERROR;
driver_name = version->name;
return AMDSMI_STATUS_SUCCESS;
}
amdsmi_status_t AMDSmiDrm::amdgpu_query_driver_date(int fd, std::string& driver_date) {
// RAII handler
using drm_version_ptr = std::unique_ptr<drmVersion,
src/amd_smi/amd_smi_gpu_device.cc
+9
Просмотреть файл
@@ -105,6 +105,15 @@ amdsmi_status_t AMDSmiGPUDevice::amdgpu_query_info(unsigned info_id,
return drm_.amdgpu_query_info(fd, info_id, size, value);
}
amdsmi_status_t AMDSmiGPUDevice::amdgpu_query_driver_name(std::string& name) const {
amdsmi_status_t ret;
uint32_t fd = 0;
ret = drm_.get_drm_fd_by_index(gpu_id_, &fd);
if (ret != AMDSMI_STATUS_SUCCESS) return AMDSMI_STATUS_NOT_SUPPORTED;
return drm_.amdgpu_query_driver_name(fd, name);
}
amdsmi_status_t AMDSmiGPUDevice::amdgpu_query_driver_date(std::string& date) const {
amdsmi_status_t ret;
uint32_t fd = 0;
src/amd_smi/amd_smi_utils.cc
+1 -1
Просмотреть файл
@@ -138,7 +138,7 @@ amdsmi_status_t smi_amdgpu_get_board_info(amd::smi::AMDSmiGPUDevice* device, amd
fp = fopen(serial_number_path.c_str(), "rb");
if (fp) {
fscanf(fp, "%lx", &info->serial_number);
fscanf(fp, "%lx", &info->product_serial);
fclose(fp);
}
tests/amd_smi_test/CMakeLists.txt
+15 -17
Просмотреть файл
@@ -7,20 +7,13 @@ set(CMAKE_INSTALL_RPATH
"\$ORIGIN:\$ORIGIN/../../../lib"
CACHE STRING "RUNPATH for tests. Helps find libgtest.so and libamd_smi.so")
# Try to find googletest
find_package(GTest 1.12.0)
# GTest_FOUND is set to TRUE if ANY version is found
# GTest_VERSION is set if 1.11.0 or newer version is found
if(NOT GTest_FOUND STREQUAL "TRUE" OR NOT DEFINED GTest_VERSION)
# Google Test wasn't found. Download and compile ourselves
include(FetchContent)
FetchContent_Declare(
googletest
GIT_REPOSITORY https://github.com/google/googletest.git
GIT_TAG release-1.12.0)
FetchContent_MakeAvailable(googletest)
endif()
# Download and compile googletest
include(FetchContent)
FetchContent_Declare(
googletest
GIT_REPOSITORY https://github.com/google/googletest.git
GIT_TAG v1.14.0)
FetchContent_MakeAvailable(googletest)
enable_testing()
@@ -70,10 +63,9 @@ target_link_libraries(${TEST}
stdc++
pthread)
# Install tests and gtest
# TODO: Remove GTest from here in the future and rely on INSTALL_GTEST?
# Install tests
install(
TARGETS ${TEST} gtest gtest_main
TARGETS ${TEST}
DESTINATION ${SHARE_INSTALL_PREFIX}/tests
COMPONENT ${TESTS_COMPONENT})
@@ -81,3 +73,9 @@ install(
FILES amdsmitst.exclude
DESTINATION ${SHARE_INSTALL_PREFIX}/tests
COMPONENT ${TESTS_COMPONENT})
# Install googletest libraries with tests
install(TARGETS gtest gtest_main
DESTINATION ${SHARE_INSTALL_PREFIX}/tests
COMPONENT ${TESTS_COMPONENT})
tests/amd_smi_test/amdsmitst.exclude
+17 -4
Просмотреть файл
@@ -14,12 +14,13 @@ PERMANENT_BLACKLIST_ALL_ASICS=
# is failing consistently
TEMPORARY_BLACKLIST_ALL_ASICS=
if [ -z $PERMANENT_BLACKLIST_ALL_ASICS -a -z $TEMPORARY_BLACKLIST_ALL_ASICS ]; then
if [ -z "$PERMANENT_BLACKLIST_ALL_ASICS" -a -z "$TEMPORARY_BLACKLIST_ALL_ASICS" ]; then
BLACKLIST_ALL_ASICS=
else
BLACKLIST_ALL_ASICS=\
"$PERMANENT_BLACKLIST_ALL_ASICS:"\
"$TEMPORARY_BLACKLIST_ALL_ASICS"
"$PERMANENT_BLACKLIST_ALL_ASICS:"\
"$TEMPORARY_BLACKLIST_ALL_ASICS:"
fi
# Device specific blacklists
@@ -55,6 +56,18 @@ FILTER[sienna_cichlid]=\
$BLACKLIST_ALL_ASICS\
"amdsmitstReadWrite.TestPerfLevelReadWrite"
# SWDEV-391407
# aqua_vanjaram and later systems show 'ip discovery' in
# /sys/class/kfd/kfd/topology/nodes/*/name
#
# For those systems gfx_target_version must be used. It can be found in
# /sys/class/kfd/kfd/topology/nodes/*/properties
FILTER[90400]=\
$BLACKLIST_ALL_ASICS\
"amdsmitstReadOnly.TestVoltCurvRead"
FILTER[90401]=${FILTER[90400]}
FILTER[90402]=${FILTER[90400]}
# SWDEV-321166
FILTER[virtualization]=\
$BLACKLIST_ALL_ASICS\
@@ -63,4 +76,4 @@ $BLACKLIST_ALL_ASICS\
"amdsmitstReadWrite.FanReadWrite:"\
"amdsmitstReadWrite.TestOverdriveReadWrite:"\
"amdsmitstReadWrite.TestPowerReadWrite:"\
"amdsmitstReadWrite.TestPowerCapReadWrite"
"amdsmitstReadWrite.TestPowerCapReadWrite"
tests/amd_smi_test/functional/frequencies_read.cc
+42 -29
Просмотреть файл
@@ -43,9 +43,7 @@
*
*/
#include <stdint.h>
#include <stddef.h>
#include <cstdint>
#include <iostream>
#include <string>
@@ -87,15 +85,23 @@ void TestFrequenciesRead::Close() {
static void print_frequencies(amdsmi_frequencies_t *f, uint32_t *l = nullptr) {
assert(f != nullptr);
for (uint32_t j = 0; j < f->num_supported; ++j) {
std::cout << "\t** " << j << ": " << f->frequency[j];
for (uint32_t clk_i = 0; clk_i < f->num_supported; ++clk_i) {
std::string clk_i_str;
if (f->has_deep_sleep) {
clk_i_str = (clk_i == 0) ? "S" : std::to_string(clk_i-1);
} else {
clk_i_str = std::to_string(clk_i);
}
std::cout << "\t** " <<
std::setw(2) << std::right << clk_i_str << ": " <<
std::setw(11) << std::right << f->frequency[clk_i];
if (l != nullptr) {
std::cout << "T/s; x" << l[j];
std::cout << "T/s; x" << l[clk_i];
} else {
std::cout << "Hz";
}
if (j == f->current) {
if (clk_i == f->current) {
std::cout << " *";
}
std::cout << std::endl;
@@ -123,23 +129,30 @@ void TestFrequenciesRead::Run(void) {
// Verify api support checking functionality is working
err = amdsmi_get_clk_freq(processor_handles_[i], t, nullptr);
ASSERT_EQ(err, AMDSMI_STATUS_INVAL);
} else if (err == AMDSMI_STATUS_NOT_YET_IMPLEMENTED) {
return;
}
if (err == AMDSMI_STATUS_NOT_YET_IMPLEMENTED) {
std::cout << "\t**Get " << name <<
": Not implemented on this machine" << std::endl;
// special driver issue, shouldn't normally occur
} else if (err == AMDSMI_STATUS_UNEXPECTED_DATA) {
return;
}
if (err == AMDSMI_STATUS_UNEXPECTED_DATA) {
// special driver issue, shouldn't normally occur
std::cerr << "WARN: Clock file [" << FreqEnumToStr(t) << "] exists on device [" << i << "] but empty!" << std::endl;
std::cerr << " Likely a driver issue!" << std::endl;
} else {
CHK_ERR_ASRT(err)
IF_VERB(STANDARD) {
std::cout << "\t**Supported " << name << " clock frequencies: ";
std::cout << f.num_supported << std::endl;
print_frequencies(&f);
// Verify api support checking functionality is working
err = amdsmi_get_clk_freq(processor_handles_[i], t, nullptr);
ASSERT_EQ(err, AMDSMI_STATUS_INVAL);
}
return;
}
CHK_ERR_ASRT(err)
IF_VERB(STANDARD) {
std::cout << "\t**Supported " << name << " clock frequencies: ";
std::cout << f.num_supported << std::endl;
print_frequencies(&f);
// Verify api support checking functionality is working
err = amdsmi_get_clk_freq(processor_handles_[i], t, nullptr);
ASSERT_EQ(err, AMDSMI_STATUS_INVAL);
}
};
@@ -162,15 +175,15 @@ void TestFrequenciesRead::Run(void) {
std::cout << "\t**Get PCIE Bandwidth "
<< ": Not implemented on this machine" << std::endl;
} else {
CHK_ERR_ASRT(err)
IF_VERB(STANDARD) {
std::cout << "\t**Supported PCIe bandwidths: ";
std::cout << b.transfer_rate.num_supported << std::endl;
print_frequencies(&b.transfer_rate, b.lanes);
// Verify api support checking functionality is working
err = amdsmi_get_gpu_pci_bandwidth(processor_handles_[i], nullptr);
ASSERT_EQ(err, AMDSMI_STATUS_INVAL);
}
CHK_ERR_ASRT(err)
IF_VERB(STANDARD) {
std::cout << "\t**Supported PCIe bandwidths: ";
std::cout << b.transfer_rate.num_supported << std::endl;
print_frequencies(&b.transfer_rate, b.lanes);
// Verify api support checking functionality is working
err = amdsmi_get_gpu_pci_bandwidth(processor_handles_[i], nullptr);
ASSERT_EQ(err, AMDSMI_STATUS_INVAL);
}
}
}
}