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

SWDEV-436531 - Changed human readable topology output to tables

Просмотр исходного кода 
Signed-off-by: Maisam Arif <maisarif@amd.com>
Change-Id: I104ff21f650ea7fd6804d3b60da9b1feddb2a701


[ROCm/amdsmi commit: 16ed186760]
Этот коммит содержится в:
Maisam Arif
2023-12-14 07:18:38 -06:00
коммит произвёл Maisam Arif
родитель 9dc60e00cb
Коммит 632443e2a3
2 изменённых файлов: 75 добавлений и 18 удалений
Скачать Patch файл Скачать Diff файл Показать все файлы Свернуть все файлы
projects/amdsmi/amdsmi_cli/amdsmi_commands.py
+59 -3
Просмотреть файл
@@ -1728,14 +1728,21 @@ class AMDSMICommands():
if not any([args.access, args.weight, args.hops, args.link_type, args.numa_bw]):
args.access = args.weight = args.hops = args.link_type= args.numa_bw = True
# Clear the table header; TODO make this a function
self.logger.table_header = ''.rjust(12)
# Populate the possible gpus
topo_values = []
for gpu in args.gpu:
gpu_id = self.helpers.get_gpu_id_from_device_handle(gpu)
topo_values.append({"gpu" : gpu_id})
gpu_bdf = amdsmi_interface.amdsmi_get_gpu_device_bdf(gpu)
self.logger.table_header += gpu_bdf.rjust(13)
if args.access:
tabular_output = []
for src_gpu_index, src_gpu in enumerate(args.gpu):
tabular_output_dict = {'gpu': amdsmi_interface.amdsmi_get_gpu_device_bdf(src_gpu)}
src_gpu_links = {}
for dest_gpu in args.gpu:
dest_gpu_id = self.helpers.get_gpu_id_from_device_handle(dest_gpu)
@@ -1756,8 +1763,18 @@ class AMDSMICommands():
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):
tabular_output_dict = {'gpu': amdsmi_interface.amdsmi_get_gpu_device_bdf(src_gpu)}
src_gpu_weight = {}
for dest_gpu in args.gpu:
dest_gpu_id = self.helpers.get_gpu_id_from_device_handle(dest_gpu)
@@ -1779,8 +1796,18 @@ class AMDSMICommands():
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):
tabular_output_dict = {'gpu': amdsmi_interface.amdsmi_get_gpu_device_bdf(src_gpu)}
src_gpu_hops = {}
for dest_gpu in args.gpu:
dest_gpu_id = self.helpers.get_gpu_id_from_device_handle(dest_gpu)
@@ -1802,8 +1829,18 @@ class AMDSMICommands():
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):
tabular_output_dict = {'gpu': amdsmi_interface.amdsmi_get_gpu_device_bdf(src_gpu)}
src_gpu_link_type = {}
for dest_gpu in args.gpu:
dest_gpu_id = self.helpers.get_gpu_id_from_device_handle(dest_gpu)
@@ -1830,8 +1867,18 @@ class AMDSMICommands():
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):
tabular_output_dict = {'gpu': amdsmi_interface.amdsmi_get_gpu_device_bdf(src_gpu)}
src_gpu_link_type = {}
for dest_gpu in args.gpu:
dest_gpu_id = self.helpers.get_gpu_id_from_device_handle(dest_gpu)
@@ -1867,6 +1914,14 @@ class AMDSMICommands():
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)
self.logger.multiple_device_output = topo_values
if self.logger.is_csv_format():
@@ -1875,7 +1930,8 @@ class AMDSMICommands():
new_output.append(self.logger.flatten_dict(elem, topology_override=True))
self.logger.multiple_device_output = new_output
self.logger.print_output(multiple_device_enabled=True)
if not self.logger.is_human_readable_format():
self.logger.print_output(multiple_device_enabled=True)
def set_value(self, args, multiple_devices=False, gpu=None, fan=None, perf_level=None,
@@ -2375,9 +2431,9 @@ class AMDSMICommands():
# Store timestamp for watch output
if watching_output:
self.logger.store_output(args.gpu, 'timestamp', int(time.time()))
self.logger.table_header += 'TIMESTAMP'.rjust(10)
self.logger.table_header += 'TIMESTAMP'.rjust(10) + ' '
self.logger.table_header += 'GPU'.rjust(3)
self.logger.table_header += 'GPU'
if args.power_usage:
try:
projects/amdsmi/amdsmi_cli/amdsmi_logger.py
+16 -15
Просмотреть файл
@@ -39,7 +39,7 @@ class AMDSMILogger():
self.format = format # csv, json, or human_readable
self.destination = destination # stdout, path to a file (append)
self.table_header = ""
self.added_table_header = False
self.table_title = ""
self.helpers = AMDSMIHelpers()
@@ -106,20 +106,18 @@ class AMDSMILogger():
value = str(value)
if key == 'gpu':
table_values += value.rjust(3)
elif key == 'timestamp':
table_values += value.rjust(10) + ' '
elif key == 'power_usage':
table_values += value.rjust(7)
elif key in ('gfx_clock', 'mem_clock', 'encoder_clock', 'decoder_clock'):
elif key in ('gfx_clock', 'mem_clock', 'encoder_clock', 'decoder_clock', 'vram_used'):
table_values += value.rjust(11)
elif key == 'throttle_status':
table_values += value.rjust(13)
elif key == 'pcie_replay':
table_values += value.rjust(13)
elif key == 'vram_total':
table_values += value.rjust(12)
elif key == 'vram_used':
table_values += value.rjust(11)
elif 'ecc' in key:
elif key == 'vram_total' or 'ecc' in key:
table_values += value.rjust(12)
elif key in ('throttle_status', 'pcie_replay'):
table_values += value.rjust(13)
elif 'gpu_' in key: # handle topology tables
table_values += value.rjust(13)
else:
table_values += value.rjust(10)
return table_values.rstrip()
@@ -391,9 +389,10 @@ class AMDSMILogger():
def _print_human_readable_output(self, multiple_device_enabled=False, watching_output=False, tabular=False):
human_readable_output = ''
if tabular and not self.added_table_header:
if tabular:
if self.table_title:
human_readable_output += self.table_title + ':\n'
human_readable_output += self.table_header + '\n'
self.added_table_header = True
if multiple_device_enabled:
for output in self.multiple_device_output:
@@ -414,12 +413,14 @@ class AMDSMILogger():
with self.destination.open('w') as output_file:
human_readable_output = ''
if tabular:
if self.table_title:
human_readable_output += self.table_title + '\n'
human_readable_output += self.table_header + '\n'
for output in self.watch_output:
human_readable_output += self._convert_json_to_human_readable(output, tabular=tabular)
if tabular:
human_readable_output += '\n'
output_file.write(human_readable_output + '\n')
else:
with self.destination.open('a') as output_file:
if tabular:
human_readable_output += self.table_header + '\n'
output_file.write(human_readable_output + '\n')