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

Added new ecc blocks and adjusted metric --ecc-block filtering

Browse Source 
Signed-off-by: Maisam Arif <maisarif@amd.com>
Change-Id: Ib2f69c7d59ee5108024794434fb202b5e4f58738


[ROCm/amdsmi commit: 1bd18c1a65]
This commit is contained in:
Maisam Arif
2024-04-12 12:43:32 -05:00
committed by Maisam Arif
parent 9b4f0f1d2b
commit cebb07e795
8 changed files with 132 additions and 50 deletions
Download Patch File Download Diff File Expand all files Collapse all files
projects/amdsmi/amdsmi_cli/amdsmi_commands.py
+3 -2
View File
@@ -1692,14 +1692,15 @@ class AMDSMICommands():
if "ecc_blocks" in current_platform_args:
if args.ecc_blocks:
ecc_dict = {}
uncountable_blocks = ["ATHUB", "DF", "SMN", "SEM", "FUSE"]
sysfs_blocks = ["UMC", "SDMA", "GFX", "MMHUB", "PCIE_BIF", "HDP", "XGMI_WAFL"]
try:
ras_states = amdsmi_interface.amdsmi_get_gpu_ras_block_features_enabled(args.gpu)
for state in ras_states:
# Only add enabled blocks that are also in sysfs
if state['status'] == amdsmi_interface.AmdSmiRasErrState.ENABLED.name:
gpu_block = amdsmi_interface.AmdSmiGpuBlock[state['block']]
# if the blocks are uncountable do not add them at all.
if gpu_block.name not in uncountable_blocks:
if gpu_block.name in sysfs_blocks:
try:
ecc_count = amdsmi_interface.amdsmi_get_gpu_ecc_count(args.gpu, gpu_block)
ecc_dict[state['block']] = {'correctable_count' : ecc_count['correctable_count'],