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

[SWDEV-495169] Update ROCm SMI CLI and Error handling (#3)

Browse Source 
Issues include:

Update ROCm SMI displaying None or Not Supported to N/A
Update ROCm SMI displaying err msg to instead log err

Signed-off-by: Juan Castillo juan.castillo@amd.com
Change-Id: I1a2ce6e4f329666b5666664a7d7b4475d6c1cbc7

[ROCm/rocm_smi_lib commit: 55ee3cc442]
This commit is contained in:
Galantsev, Dmitrii
2025-01-14 17:15:18 -06:00
committed by GitHub
parent 84400150b4
commit bc13dfe3c8
2 changed files with 15 additions and 7 deletions
Download Patch File Download Diff File Expand all files Collapse all files
projects/rocm-smi-lib/python_smi_tools/rocm_smi.py
+9 -3
View File
@@ -1981,22 +1981,28 @@ def showAllConcise(deviceList):
+ getComputePartition(device, silent) + getComputePartition(device, silent)
+ ", " + getPartitionId(device, silent)) + ", " + getPartitionId(device, silent))
sclk = showCurrentClocks([device], 'sclk', concise=silent) sclk = showCurrentClocks([device], 'sclk', concise=silent)
if not sclk:
sclk = 'N/A'
mclk = showCurrentClocks([device], 'mclk', concise=silent) mclk = showCurrentClocks([device], 'mclk', concise=silent)
if not mclk:
mclk = 'N/A'
(retCode, fanLevel, fanSpeed) = getFanSpeed(device, silent) (retCode, fanLevel, fanSpeed) = getFanSpeed(device, silent)
fan = str(fanSpeed) + '%' fan = str(fanSpeed) + '%'
if getPerfLevel(device, silent) != -1: if getPerfLevel(device, silent) != -1:
perf = getPerfLevel(device, silent) perf = getPerfLevel(device, silent)
else: else:
perf = 'Unsupported' perf = 'N/A'
if getMaxPower(device, silent) != -1: if getMaxPower(device, silent) != -1:
pwrCap = str(getMaxPower(device, silent)) + 'W' pwrCap = str(getMaxPower(device, silent)) + 'W'
else: else:
pwrCap = 'Unsupported' pwrCap = 'N/A'
if getGpuUse(device, silent) != -1: if getGpuUse(device, silent) != -1:
gpu_busy = str(getGpuUse(device, silent)) + '%' gpu_busy = str(getGpuUse(device, silent)) + '%'
else: else:
gpu_busy = 'Unsupported' gpu_busy = 'N/A'
allocated_mem_percent = getAllocatedMemoryPercent(device) allocated_mem_percent = getAllocatedMemoryPercent(device)
if allocated_mem_percent['ret'] != rsmi_status_t.RSMI_STATUS_SUCCESS:
allocated_mem_percent['combined'] = 'N/A'
# Top Row - per device data # Top Row - per device data
values['card%s' % (str(device))] = [device, getNodeId(device), values['card%s' % (str(device))] = [device, getNodeId(device),
projects/rocm-smi-lib/src/rocm_smi.cc
+6 -4
View File
@@ -368,8 +368,9 @@ static rsmi_status_t get_dev_mon_value(amd::smi::MonitorTypes type,
} }
if (!amd::smi::IsInteger(val_str)) { if (!amd::smi::IsInteger(val_str)) {
std::cerr << "Expected integer value from monitor," std::ostringstream ss;
" but got \"" << val_str << "\"" << std::endl; ss << "Expected integer value from monitor, but got \"" << val_str << "\"";
LOG_ERROR(ss);
return RSMI_STATUS_UNEXPECTED_DATA; return RSMI_STATUS_UNEXPECTED_DATA;
} }
@@ -397,8 +398,9 @@ static rsmi_status_t get_dev_mon_value(amd::smi::MonitorTypes type,
} }
if (!amd::smi::IsInteger(val_str)) { if (!amd::smi::IsInteger(val_str)) {
std::cerr << "Expected integer value from monitor," std::ostringstream ss;
" but got \"" << val_str << "\"" << std::endl; ss << "Expected integer value from monitor, but got \"" << val_str << "\"";
LOG_ERROR(ss);
return RSMI_STATUS_UNEXPECTED_DATA; return RSMI_STATUS_UNEXPECTED_DATA;
} }