[SWDEV-490187] reset gpu partition were removed

The reset gpu partition support for both compute and memory were removed

Code changes related to the following:
  * amdsmi_reset_gpu_compute_partition()
  * amdsmi_reset_gpu_memory_partition()
  * CLI

Change-Id: I372589074b4da172bedd39223edde18939e373ae
Signed-off-by: Oliveira, Daniel <daniel.oliveira@amd.com>
Цей коміт міститься в:
gabrpham committed 2024-10-18 16:22:26 -05:00
1 джерело 2e5b164c43
коміт f5b7761ac7
10 files changed
+79 -273

No files matched your search

+43 -68
Переглянути файл
@@ -4210,7 +4210,7 @@ class AMDSMICommands():
def reset(self, args, multiple_devices=False, gpu=None, gpureset=None,
clocks=None, fans=None, profile=None, xgmierr=None, perf_determinism=None,
compute_partition=None, memory_partition=None, power_cap=None, clean_local_data=None):
power_cap=None, clean_local_data=None):
"""Issue reset commands to target gpu(s)
Args:
@@ -4223,8 +4223,6 @@ class AMDSMICommands():
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.
compute_partition (bool, optional): Value override for args.compute_partition. Defaults to None.
memory_partition (bool, optional): Value override for args.memory_partition. 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.
@@ -4250,10 +4248,6 @@ class AMDSMICommands():
args.xgmierr = xgmierr
if perf_determinism:
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 clean_local_data:
@@ -4355,67 +4349,48 @@ class AMDSMICommands():
reset_profile_results['power_profile'] = "N/A"
logging.debug("Failed to reset power profile on gpu %s | %s", gpu_id, e.get_error_info())
try:
level_auto = amdsmi_interface.AmdSmiDevPerfLevel.AUTO
amdsmi_interface.amdsmi_set_gpu_perf_level(args.gpu, level_auto)
reset_profile_results['performance_level'] = 'Successfully reset Performance 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
reset_profile_results['performance_level'] = "N/A"
logging.debug("Failed to reset perf level on gpu %s | %s", gpu_id, e.get_error_info())
self.logger.store_output(args.gpu, 'reset_profile', reset_profile_results)
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
result = "N/A"
logging.debug("Failed to reset xgmi error count on gpu %s | %s", gpu_id, e.get_error_info())
self.logger.store_output(args.gpu, 'reset_xgmi_err', result)
if args.perf_determinism:
try:
level_auto = amdsmi_interface.AmdSmiDevPerfLevel.AUTO
amdsmi_interface.amdsmi_set_gpu_perf_level(args.gpu, level_auto)
result = 'Successfully disabled performance 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
result = "N/A"
logging.debug("Failed to set perf level on gpu %s | %s", gpu_id, e.get_error_info())
self.logger.store_output(args.gpu, 'reset_perf_determinism', result)
if args.compute_partition:
try:
amdsmi_interface.amdsmi_reset_gpu_compute_partition(args.gpu)
result = 'Successfully reset compute partition'
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 = "N/A"
logging.debug("Failed to reset compute partition on gpu %s | %s", gpu_id, e.get_error_info())
self.logger.store_output(args.gpu, 'reset_compute_partition', result)
if args.memory_partition:
try:
amdsmi_interface.amdsmi_reset_gpu_memory_partition(args.gpu)
result = 'Successfully reset memory partition'
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 = "N/A"
logging.debug("Failed to reset memory partition on gpu %s | %s", gpu_id, e.get_error_info())
self.logger.store_output(args.gpu, 'reset_memory_partition', result)
if args.power_cap:
try:
power_cap_info = amdsmi_interface.amdsmi_get_power_cap_info(args.gpu)
logging.debug(f"Power cap info for gpu {gpu_id} | {power_cap_info}")
default_power_cap_in_w = power_cap_info["default_power_cap"]
default_power_cap_in_w = self.helpers.convert_SI_unit(default_power_cap_in_w, AMDSMIHelpers.SI_Unit.MICRO)
current_power_cap_in_w = power_cap_info["power_cap"]
current_power_cap_in_w = self.helpers.convert_SI_unit(current_power_cap_in_w, AMDSMIHelpers.SI_Unit.MICRO)
except amdsmi_exception.AmdSmiLibraryException as e:
raise ValueError(f"Unable to get power cap info from {gpu_id}") from e
try:
level_auto = amdsmi_interface.AmdSmiDevPerfLevel.AUTO
amdsmi_interface.amdsmi_set_gpu_perf_level(args.gpu, level_auto)
reset_profile_results['performance_level'] = 'Successfully reset Performance 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
reset_profile_results['performance_level'] = "N/A"
logging.debug("Failed to reset perf level on gpu %s | %s", gpu_id, e.get_error_info())
self.logger.store_output(args.gpu, 'reset_profile', reset_profile_results)
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
result = "N/A"
logging.debug("Failed to reset xgmi error count on gpu %s | %s", gpu_id, e.get_error_info())
self.logger.store_output(args.gpu, 'reset_xgmi_err', result)
if args.perf_determinism:
try:
level_auto = amdsmi_interface.AmdSmiDevPerfLevel.AUTO
amdsmi_interface.amdsmi_set_gpu_perf_level(args.gpu, level_auto)
result = 'Successfully disabled performance 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
result = "N/A"
logging.debug("Failed to set perf level on gpu %s | %s", gpu_id, e.get_error_info())
self.logger.store_output(args.gpu, 'reset_perf_determinism', result)
if args.power_cap:
try:
power_cap_info = amdsmi_interface.amdsmi_get_power_cap_info(args.gpu)
logging.debug(f"Power cap info for gpu {gpu_id} | {power_cap_info}")
default_power_cap_in_w = power_cap_info["default_power_cap"]
default_power_cap_in_w = self.helpers.convert_SI_unit(default_power_cap_in_w, AMDSMIHelpers.SI_Unit.MICRO)
current_power_cap_in_w = power_cap_info["power_cap"]
current_power_cap_in_w = self.helpers.convert_SI_unit(current_power_cap_in_w, AMDSMIHelpers.SI_Unit.MICRO)
except amdsmi_exception.AmdSmiLibraryException as e:
raise ValueError(f"Unable to get power cap info from {gpu_id}") from e
if current_power_cap_in_w == default_power_cap_in_w:
self.logger.store_output(args.gpu, 'powercap', f"Power cap is already set to {default_power_cap_in_w}")
-4
Переглянути файл
@@ -1128,8 +1128,6 @@ class AMDSMIParser(argparse.ArgumentParser):
reset_profile_help = "Reset power profile back to default"
reset_xgmierr_help = "Reset XGMI error counts"
reset_perf_det_help = "Disable performance determinism"
reset_compute_help = "Reset compute partitions on the specified GPU"
reset_memory_help = "Reset memory partitions on the specified GPU"
reset_power_cap_help = "Reset power capacity limit to max capable"
reset_gpu_clean_local_data_help = "Clean up local data in LDS/GPRs on a per partition basis"
@@ -1152,8 +1150,6 @@ class AMDSMIParser(argparse.ArgumentParser):
reset_parser.add_argument('-p', '--profile', action='store_true', required=False, help=reset_profile_help)
reset_parser.add_argument('-x', '--xgmierr', action='store_true', required=False, help=reset_xgmierr_help)
reset_parser.add_argument('-d', '--perf-determinism', action='store_true', required=False, help=reset_perf_det_help)
reset_parser.add_argument('-C', '--compute-partition', action='store_true', required=False, help=reset_compute_help)
reset_parser.add_argument('-M', '--memory-partition', action='store_true', required=False, help=reset_memory_help)
reset_parser.add_argument('-o', '--power-cap', action='store_true', required=False, help=reset_power_cap_help)
# Add Baremetal and Virtual OS reset arguments