AI/rocm-systems
2
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Removed uncountable blocks from metric --ecc-block

Browse Source 
Signed-off-by: Maisam Arif <maisarif@amd.com>
Change-Id: Iab8ed3f80c8ebd7ffd3bfcf197e1a0b03782b498
This commit is contained in:
Maisam Arif
2023-10-10 10:32:20 -05:00
parent 6ca95c1a2d
commit b47d028da2
1 changed files with 11 additions and 8 deletions
Download Patch File Download Diff File Expand all files Collapse all files
amdsmi_cli/amdsmi_commands.py
+11 -8
View File
@@ -1003,19 +1003,22 @@ class AMDSMICommands():
values_dict['ecc'] = ecc_count
if args.ecc_block:
ecc_dict = {}
uncountable_blocks = ["ATHUB", "DF", "SMN", "SEM", "MP0", "MP1", "FUSE"]
try:
ras_states = amdsmi_interface.amdsmi_get_gpu_ras_block_features_enabled(args.gpu)
for state in ras_states:
if state['status'] == amdsmi_interface.AmdSmiRasErrState.ENABLED.name:
gpu_block = amdsmi_interface.AmdSmiGpuBlock[state['block']]
try:
ecc_count = amdsmi_interface.amdsmi_get_gpu_ecc_count(args.gpu, gpu_block)
ecc_dict[state['block']] = {'correctable' : ecc_count['correctable_count'],
'uncorrectable': ecc_count['uncorrectable_count']}
except amdsmi_exception.AmdSmiLibraryException as e:
ecc_dict[state['block']] = {'correctable' : "N/A",
'uncorrectable': "N/A"}
logging.debug("Failed to get ecc count for gpu %s at block %s | %s", gpu_id, gpu_block, e.get_error_info())
# if the blocks are uncountable do not add them at all.
if gpu_block.name not in uncountable_blocks:
try:
ecc_count = amdsmi_interface.amdsmi_get_gpu_ecc_count(args.gpu, gpu_block)
ecc_dict[state['block']] = {'correctable' : ecc_count['correctable_count'],
'uncorrectable': ecc_count['uncorrectable_count']}
except amdsmi_exception.AmdSmiLibraryException as e:
ecc_dict[state['block']] = {'correctable' : "N/A",
'uncorrectable': "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_block'] = ecc_dict
except amdsmi_exception.AmdSmiLibraryException as e: