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

[SWDEV-423481/SWDEV-423393] Align all device identifier details

Browse Source 
Updated:
 * [CLI] Fixed vram % - printf style formatting causes many data errors
   This fix updates to the recommended way of outputting formatted data.
   https://docs.python.org/3/library/stdtypes.html#printf-style-string-formatting
 * [API/CLI] Added gpu_id / GUID from kfd (rsmi_dev_guid_get)
       -> CLI name: "GUID"
       -> ROCm SMI calls: no arg, -i, --showhw, --showproduct
 * [API/CLI] Added node_id from kfd (rsmi_dev_node_get)
       -> CLI name: "Node"
       -> ROCm SMI calls: no arg, --showhw, --showproduct
 * [CLI] Added target gfx version from kfd
       -> CLI name: "GFX Version" or "GFX VER"
       -> ROCm SMI calls: --showhw, --showproduct
 * [CLI] Base ROCm CLI
       -> Removed - stacked id formatting:
	   This is to simplify identifiers helpful to users.
	   More identifiers can be found on -i --showhw, --showproduct
 * [CLI] Update -i, --showhw, --showproduct, w/out arg
      -> Card ID/DID/Model/SKU/VBIOS:
            All unsupported values now display "N/A" instead
            of "unknown" or "unsupported"
 * [CLI] Showhw now expands data based on content

Change-Id: Ifb8586f9f545892b8a5aa7903608273cdd77e075
Signed-off-by: Charis Poag <Charis.Poag@amd.com>
This commit is contained in:
Charis Poag
2024-01-31 21:03:33 -06:00
parent 677433b367
commit 4b5ccb57f0
8 changed files with 403 additions and 171 deletions
Download Patch File Download Diff File Expand all files Collapse all files
python_smi_tools/rocm_smi.py
+181 -160
View File
@@ -254,7 +254,7 @@ def getGpuUse(device, silent=False):
return -1
def getId(device, silent=False):
def getDRMDeviceId(device, silent=False):
""" Return the hexadecimal value of a device's ID
@param device: DRM device identifier
@@ -263,8 +263,10 @@ def getId(device, silent=False):
"""
dv_id = c_short()
ret = rocmsmi.rsmi_dev_id_get(device, byref(dv_id))
device_id_ret = "N/A"
if rsmi_ret_ok(ret, device, 'get_device_id', silent):
return hex(dv_id.value)
device_id_ret = hex(dv_id.value)
return device_id_ret
def getRev(device, silent=False):
@@ -276,9 +278,103 @@ def getRev(device, silent=False):
"""
dv_rev = c_short()
ret = rocmsmi.rsmi_dev_revision_get(device, byref(dv_rev))
if rsmi_ret_ok(ret, device, 'get_device_rev', silent):
return hex(dv_rev.value)
revision_ret = "N/A"
if rsmi_ret_ok(ret, device, 'get_device_rev', silent=silent):
revision_ret = padHexValue(hex(dv_rev.value), 2)
return revision_ret
def getSubsystemId(device, silent=False):
""" Return the a device's subsystem id
@param device: DRM device identifier
@param silent=Turn on to silence error output
(you plan to handle manually). Default is off.
"""
model = create_string_buffer(MAX_BUFF_SIZE)
ret = rocmsmi.rsmi_dev_subsystem_name_get(device, model, MAX_BUFF_SIZE)
device_model = "N/A"
if rsmi_ret_ok(ret, device, 'get_subsystem_name', silent=silent):
device_model = model.value.decode()
# padHexValue is used for applications that expect 4-digit card models
device_model = padHexValue(device_model, 4)
return device_model
def getVendor(device, silent=False):
""" Return the a device's vendor id
@param device: DRM device identifier
@param silent=Turn on to silence error output
(you plan to handle manually). Default is off.
"""
vendor = create_string_buffer(MAX_BUFF_SIZE)
device_vendor = "N/A"
# Retrieve card vendor
ret = rocmsmi.rsmi_dev_vendor_name_get(device, vendor, MAX_BUFF_SIZE)
# Only continue if GPU vendor is AMD
if rsmi_ret_ok(ret, device, 'get_vendor_name', silent) and isAmdDevice(device):
device_vendor = vendor.value.decode()
return device_vendor
def getGUID(device, silent=False):
""" Return the uint64 value of device's GUID,
also referred as GPU ID - reported by KFD.
@param device: DRM device identifier
@param silent=Turn on to silence error output
(you plan to handle manually). Default is off.
"""
guid = c_uint64()
ret = rocmsmi.rsmi_dev_guid_get(device, byref(guid))
guid_ret = "N/A"
if rsmi_ret_ok(ret, device, 'get_gpu_id_kfd', silent=silent):
guid_ret = guid.value
return guid_ret
def getTargetGfxVersion(device, silent=False):
""" Return the uint64 value of device's target
graphics version as reported by KFD
@param device: DRM device identifier
@param silent=Turn on to silence error output
(you plan to handle manually). Default is off.
"""
gfx_version = c_uint64()
gfx_ver_ret = "N/A"
ret = rocmsmi.rsmi_dev_target_graphics_version_get(device, byref(gfx_version))
if rsmi_ret_ok(ret, device, 'get_target_gfx_version', silent=silent):
gfx_ver_ret = "gfx" + str(gfx_version.value)
return gfx_ver_ret
def getNodeId(device, silent=False):
""" Return the uint32 value of device's node id
reported by KFD.
@param device: DRM device identifier
@param silent=Turn on to silence error output
(you plan to handle manually). Default is off.
"""
node_id = c_uint32()
ret = rocmsmi.rsmi_dev_node_id_get(device, byref(node_id))
node_id_ret = "N/A"
if rsmi_ret_ok(ret, device, 'get_node_id_kfd', silent=silent):
node_id_ret = node_id.value
return node_id_ret
def getDeviceName(device, silent=False):
""" Return the uint64 value of device's target
graphics version as reported by KFD
@param device: DRM device identifier
@param silent=Turn on to silence error output
(you plan to handle manually). Default is off.
"""
# Retrieve the device series
series = create_string_buffer(MAX_BUFF_SIZE)
device_name_ret = "N/A"
ret = rocmsmi.rsmi_dev_name_get(device, series, MAX_BUFF_SIZE)
if rsmi_ret_ok(ret, device, 'get_name', silent=silent):
device_name_ret = series.value.decode()
return device_name_ret
def getMaxPower(device, silent=False):
""" Return the maximum power cap of a given device
@@ -515,10 +611,12 @@ def getVbiosVersion(device, silent=False):
"""
vbios = create_string_buffer(256)
ret = rocmsmi.rsmi_dev_vbios_version_get(device, vbios, 256)
if ret == rsmi_status_t.RSMI_STATUS_NOT_SUPPORTED:
return "Unsupported"
elif rsmi_ret_ok(ret, device, silent=silent):
return vbios.value.decode()
vbios_ret = "N/A"
if rsmi_ret_ok(ret, device, silent=silent):
vbios_ret = vbios.value.decode()
if vbios_ret == "":
vbios_ret = "N/A"
return vbios_ret
def getVersion(deviceList, component, silent=False):
@@ -1785,9 +1883,9 @@ def showAllConcise(deviceList):
deviceList.sort()
available_temp_type = getTemperatureLabel(deviceList)
temp_type = "(" + available_temp_type.capitalize() + ")"
header=['Device', '[Model : Revision]', 'Temp', 'Power', 'Partitions',
header=['Device', 'Node','IDs','', 'Temp', 'Power', 'Partitions',
'SCLK', 'MCLK', 'Fan', 'Perf', 'PwrCap', 'VRAM%', 'GPU%']
subheader = ['', 'Name (20 chars)', temp_type, getPowerLabel(deviceList),
subheader = ['', '','(DID, ', 'GUID)', temp_type, getPowerLabel(deviceList),
'(Mem, Compute)', '', '', '', '', '', '', '']
# add additional spaces to match header
for idx, item in enumerate(subheader):
@@ -1805,8 +1903,6 @@ def showAllConcise(deviceList):
values = {}
degree_sign = u'\N{DEGREE SIGN}'
for device in deviceList:
gpu_dev_product_info = getDevProductInfo(device, silent)
gpu_dev_product_info_names = list(gpu_dev_product_info[device])
temp_val = str(getTemp(device, available_temp_type, silent))
if temp_val != 'N/A':
temp_val += degree_sign + 'C'
@@ -1839,19 +1935,19 @@ def showAllConcise(deviceList):
if vram_used is None:
mem_use_pct='Unsupported'
if vram_used != None and vram_total != None and float(vram_total) != 0:
mem_use_pct = '% 3.0f%%' % (100 * (float(vram_used) / float(vram_total)))
mem_use_pct = float(100 * (float(vram_used) / float(vram_total)))
mem_use_pct = '{:<.0f}%'.format(mem_use_pct) # left aligned
# values with no precision
gpu_dev_product_info_top_name = gpu_dev_product_info_names[0]
if (len(gpu_dev_product_info_names) > 1):
values['card%s_Info' % (str(device))] = ['', gpu_dev_product_info_names[0], '', '', '',
'', '', '',
'', '', '', '']
gpu_dev_product_info_top_name = gpu_dev_product_info_names[1]
values['card%s' % (str(device))] = [device, gpu_dev_product_info_top_name, temp_val,
powerVal, combined_partition, sclk, mclk,
fan, str(perf).lower(), pwrCap, mem_use_pct,
gpu_busy]
# Top Row - per device data
values['card%s' % (str(device))] = [device, getNodeId(device),
str(getDRMDeviceId(device)) + ", ",
str(getGUID(device)),
temp_val, powerVal, combined_partition,
sclk, mclk, fan, str(perf).lower(),
str(pwrCap),
str(mem_use_pct),
str(gpu_busy)]
val_widths = {}
for device in deviceList:
@@ -1875,18 +1971,13 @@ def showAllConcise(deviceList):
for device in deviceList:
printLog(None, "".join(str(word).ljust(max_widths[col]) for col, word in
zip(range(len(max_widths)), values['card%s' % (str(device))])), None)
gpu_dev_product_info = getDevProductInfo(device, silent)
gpu_dev_product_info_names = list(gpu_dev_product_info[device])
if (len(gpu_dev_product_info_names) > 1):
printLog(None, "".join(str(word).ljust(max_widths[col]) for col, word in
zip(range(len(max_widths)), values['card%s_Info' % (str(device))])), None)
printLogSpacer(contentSizeToFit=len(header_output))
printLogSpacer(footerString, contentSizeToFit=len(header_output))
def showAllConciseHw(deviceList):
""" Display critical Hardware info for all devices in a concise format
""" Display critical Hardware info
@param deviceList: List of DRM devices (can be a single-item list)
"""
@@ -1894,25 +1985,22 @@ def showAllConciseHw(deviceList):
if PRINT_JSON:
print('ERROR: Cannot print JSON/CSV output for concise hardware output')
sys.exit(1)
printLogSpacer(' Concise Hardware Info ')
header = ['GPU', 'DID', 'DREV', 'GFX RAS', 'SDMA RAS', 'UMC RAS', 'VBIOS', 'BUS']
header = ['GPU', 'NODE', 'DID', 'GUID', 'GFX VER', 'GFX RAS', 'SDMA RAS', 'UMC RAS', 'VBIOS', 'BUS']
head_widths = [len(head) + 2 for head in header]
values = {}
silent = True
for device in deviceList:
gpuid = getId(device, silent)
if str(gpuid).startswith('0x'):
gpuid = str(gpuid)[2:]
gpurev = getRev(device, silent)
if str(gpurev).startswith('0x'):
gpurev = str(gpurev)[2:]
did = getDRMDeviceId(device, silent)
nodeid = getNodeId(device, silent)
guid = getGUID(device, silent)
gfxVer = getTargetGfxVersion(device, silent)
gfxRas = getRasEnablement(device, 'GFX', silent)
sdmaRas = getRasEnablement(device, 'SDMA', silent)
umcRas = getRasEnablement(device, 'UMC', silent)
vbios = getVbiosVersion(device, silent)
bus = getBus(device, silent)
values['card%s' % (str(device))] = [device, gpuid, gpurev, gfxRas, sdmaRas, umcRas, vbios, bus]
values['card%s' % (str(device))] = [device, nodeid, did, guid, gfxVer, gfxRas, sdmaRas,
umcRas, vbios, bus]
val_widths = {}
for device in deviceList:
val_widths[device] = [len(str(val)) + 2 for val in values['card%s' % (str(device))]]
@@ -1920,11 +2008,25 @@ def showAllConciseHw(deviceList):
for device in deviceList:
for col in range(len(val_widths[device])):
max_widths[col] = max(max_widths[col], val_widths[device][col])
printLog(None, "".join(word.ljust(max_widths[col]) for col, word in zip(range(len(max_widths)), header)), None)
device_output=""
for device in deviceList:
printLog(None, "".join(str(word).ljust(max_widths[col]) for col, word in
zip(range(len(max_widths)), values['card%s' % (str(device))])), None)
printLogSpacer()
if (device + 1 != len(deviceList)):
device_output += "".join(str(word).ljust(max_widths[col]) for col, word in
zip(range(len(max_widths)), values['card%s' % (str(device))])) + "\n"
else:
device_output += "".join(str(word).ljust(max_widths[col]) for col, word in
zip(range(len(max_widths)), values['card%s' % (str(device))]))
#################################
# Display concise hardware info #
#################################
header_output = "".join(word.ljust(max_widths[col]) for col, word in zip(range(len(max_widths)), header))
printLogSpacer(headerString, contentSizeToFit=len(header_output))
printLogSpacer(' Concise Hardware Info ', contentSizeToFit=len(header_output))
printLog(None, header_output, None)
printLog(None, device_output, None)
printLogSpacer(fill='=', contentSizeToFit=len(header_output))
printLogSpacer(footerString, contentSizeToFit=len(header_output))
def showBus(deviceList):
@@ -2276,14 +2378,17 @@ def showEnergy(deviceList):
def showId(deviceList):
""" Display the device ID for a list of devices
""" Display the device IDs for a list of devices
@param deviceList: List of DRM devices (can be a single-item list)
"""
printLogSpacer(' ID ')
for device in deviceList:
printLog(device, 'Device ID', getId(device))
printLog(device, 'Device Rev', getRev(device))
printLog(device, 'Device Name', '\t\t' + str(getDeviceName(device)))
printLog(device, 'Device ID', '\t\t' + str(getDRMDeviceId(device)))
printLog(device, 'Device Rev', '\t\t' + str(getRev(device)))
printLog(device, 'Subsystem ID', '\t' + str(getSubsystemId(device)))
printLog(device, 'GUID', '\t\t' + str(getGUID(device)))
printLogSpacer()
@@ -2582,126 +2687,41 @@ def showPowerPlayTable(deviceList):
printLogSpacer()
def showProductName(deviceList):
""" Show the requested product name for a list of devices
def showProduct(deviceList):
""" Show the requested product information for a list of devices
@param deviceList: List of DRM devices (can be a single-item list)
"""
series = create_string_buffer(256)
model = create_string_buffer(256)
vendor = create_string_buffer(256)
vbios = create_string_buffer(256)
# sku = create_string_buffer(256)
printLogSpacer(' Product Info ')
for device in deviceList:
# Retrieve card vendor
ret = rocmsmi.rsmi_dev_vendor_name_get(device, vendor, 256)
# Only continue if GPU vendor is AMD
if rsmi_ret_ok(ret, device, 'get_vendor_name') and isAmdDevice(device):
try:
device_vendor = vendor.value.decode()
except UnicodeDecodeError:
printErrLog(device, "Unable to read device vendor")
device_vendor = "N/A"
# Retrieve the device series
ret = rocmsmi.rsmi_dev_name_get(device, series, 256)
if rsmi_ret_ok(ret, device, 'get_name'):
try:
device_series = series.value.decode()
printLog(device, 'Card series', '\t\t' + device_series)
except UnicodeDecodeError:
printErrLog(device, "Unable to read card series")
# Retrieve the device model
ret = rocmsmi.rsmi_dev_subsystem_name_get(device, model, 256)
if rsmi_ret_ok(ret, device, 'get_subsystem_name'):
try:
device_model = model.value.decode()
# padHexValue is used for applications that expect 4-digit card models
printLog(device, 'Card model', '\t\t' + padHexValue(device_model, 4))
except UnicodeDecodeError:
printErrLog(device, "Unable to read device model")
printLog(device, 'Card vendor', '\t\t' + device_vendor)
if isAmdDevice(device):
# TODO: Retrieve the SKU using 'rsmi_dev_sku_get' from the LIB
# ret = rocmsmi.rsmi_dev_sku_get(device, sku, 256)
# if rsmi_ret_ok(ret, device) and sku.value.decode():
# device_sku = sku.value.decode()
# Retrieve the device SKU as a substring from VBIOS
device_sku = ""
ret = rocmsmi.rsmi_dev_vbios_version_get(device, vbios, 256)
if ret == rsmi_status_t.RSMI_STATUS_NOT_SUPPORTED:
device_sku = "Unsupported"
printLog(device, 'Card SKU', '\t\t' + device_sku)
elif rsmi_ret_ok(ret, device, 'get_vbios_version') and vbios.value.decode():
# Device SKU is just the characters in between the two '-' in vbios_version
if vbios.value.decode().count('-') == 2 and len(str(vbios.value.decode().split('-')[1])) > 1:
device_sku = vbios.value.decode().split('-')[1]
else:
device_sku = 'unknown'
printLog(device, 'Card SKU', '\t\t' + device_sku)
else:
printErrLog(device, "Unable to decode VBIOS value for device SKU")
# Device SKU is just the characters in between the two '-' in vbios_version
vbios = getVbiosVersion(device, True)
device_sku = "N/A"
if vbios.count('-') == 2 and len(str(vbios.split('-')[1])) > 1:
device_sku = vbios.split('-')[1]
printLog(device, 'Card Series', '\t\t' + str(getDeviceName(device)))
# Retrieve device ID from DRM and KFD
printLog(device, 'Card Model', str('\t\t' + getDRMDeviceId(device)))
printLog(device, 'Card Vendor', '\t\t' + getVendor(device))
printLog(device, 'Card SKU', '\t\t' + device_sku)
printLog(device, 'Subsystem ID', str('\t' + getSubsystemId(device)))
printLog(device, 'Device Rev', str('\t\t' + getRev(device)))
printLog(device, 'Node ID', str('\t\t' + str(getNodeId(device))))
printLog(device, 'GUID', str('\t\t' + str(getGUID(device))))
printLog(device, 'GFX Version', str('\t\t' + getTargetGfxVersion(device)))
else:
vendor = getVendor(device)
printLog(device, 'Incompatible device.\n' \
'GPU[%s]\t\t: Expected vendor name: Advanced Micro Devices, Inc. [AMD/ATI]\n' \
'GPU[%s]\t\t: Actual vendor name' % (device, device), vendor.value.decode())
'GPU[%s]\t\t: Actual vendor name' % (device, device), vendor)
printLogSpacer()
def getDevProductInfo(device, silent=False):
""" Show the requested product name for the device requested
@param device: Device we want to get the info for
@param silent=Turn on to silence error output
(you plan to handle manually). Default is off.
"""
# Retrieve card vendor
MAX_DESC_SIZE = 20
device_series = "N/A"
device_model = "N/A"
gpu_revision = "N/A"
device_list = {}
vendor = create_string_buffer(MAX_BUFF_SIZE)
ret = rocmsmi.rsmi_dev_vendor_name_get(device, vendor, MAX_BUFF_SIZE)
# Only continue if GPU vendor is AMD
if rsmi_ret_ok(ret, device, 'get_vendor_name', silent) and isAmdDevice(device):
# Retrieve the device series
series = create_string_buffer(MAX_BUFF_SIZE)
ret = rocmsmi.rsmi_dev_name_get(device, series, MAX_BUFF_SIZE)
if rsmi_ret_ok(ret, device, 'get_name', silent):
try:
device_series = series.value.decode()
except UnicodeDecodeError:
if not silent:
printErrLog(device, "Unable to read card series")
# Retrieve the device model
model = create_string_buffer(MAX_BUFF_SIZE)
ret = rocmsmi.rsmi_dev_subsystem_name_get(device, model, MAX_BUFF_SIZE)
if rsmi_ret_ok(ret, device, 'get_subsystem_name', silent):
try:
device_model = model.value.decode()
device_model = padHexValue(device_model, 4)
except UnicodeDecodeError:
if not silent:
printErrLog(device, "Unable to read device model")
try:
gpu_revision = padHexValue(getRev(device), 2)
except Exception as exc:
if not silent:
printErrLog(device, "Unable to read card revision %s" % (exc))
device_series_str = str(device_series[:MAX_DESC_SIZE])
device_series_str = device_series_str.ljust(MAX_DESC_SIZE, ' ')
device_model_str = str(('[' + device_model + ' : ' + gpu_revision + ']'))
device_model_str = str(device_model_str[:MAX_DESC_SIZE])
device_model_str = device_model_str.ljust(MAX_DESC_SIZE, ' ')
device_list = {device : [device_series_str, device_model_str]}
return device_list
def showProfile(deviceList):
""" Display available Power Profiles for a list of devices.
@@ -3713,9 +3733,10 @@ def save(deviceList, savefilepath):
# The code below is for when this script is run as an executable instead of when imported as a module
def isConciseInfoRequested(args):
return len(sys.argv) == 1 or \
is_concise_req = len(sys.argv) == 1 or \
len(sys.argv) == 2 and (args.alldevices or (args.json or args.csv)) or \
len(sys.argv) == 3 and (args.alldevices and (args.json or args.csv))
return is_concise_req
if __name__ == '__main__':
parser = argparse.ArgumentParser(
@@ -3741,7 +3762,7 @@ if __name__ == '__main__':
groupDisplayOpt.add_argument('--showhw', help='Show Hardware details', action='store_true')
groupDisplayOpt.add_argument('-a', '--showallinfo', help='Show Temperature, Fan and Clock values',
action='store_true')
groupDisplayTop.add_argument('-i', '--showid', help='Show DEVICE ID', action='store_true')
groupDisplayTop.add_argument('-i', '--showid', help='Show DEVICE IDs', action='store_true')
groupDisplayTop.add_argument('-v', '--showvbios', help='Show VBIOS version', action='store_true')
groupDisplayTop.add_argument('-e', '--showevents', help='Show event list', metavar='EVENT', type=str, nargs='*')
groupDisplayTop.add_argument('--showdriverversion', help='Show kernel driver version', action='store_true')
@@ -3750,7 +3771,7 @@ if __name__ == '__main__':
groupDisplayTop.add_argument('--showmclkrange', help='Show mclk range', action='store_true')
groupDisplayTop.add_argument('--showmemvendor', help='Show GPU memory vendor', action='store_true')
groupDisplayTop.add_argument('--showsclkrange', help='Show sclk range', action='store_true')
groupDisplayTop.add_argument('--showproductname', help='Show SKU/Vendor name', action='store_true')
groupDisplayTop.add_argument('--showproductname', help='Show product details', action='store_true')
groupDisplayTop.add_argument('--showserial', help='Show GPU\'s Serial Number', action='store_true')
groupDisplayTop.add_argument('--showuniqueid', help='Show GPU\'s Unique ID', action='store_true')
groupDisplayTop.add_argument('--showvoltagerange', help='Show voltage range', action='store_true')
@@ -3933,7 +3954,7 @@ if __name__ == '__main__':
if not PRINT_JSON:
print('\n')
if not isConciseInfoRequested(args):
if not isConciseInfoRequested(args) and args.showhw == False:
printLogSpacer(headerString)
if args.showallinfo:
@@ -4060,7 +4081,7 @@ if __name__ == '__main__':
if args.showfwinfo or str(args.showfwinfo) == '[]':
showFwInfo(deviceList, args.showfwinfo)
if args.showproductname:
showProductName(deviceList)
showProduct(deviceList)
if args.showxgmierr:
showXgmiErr(deviceList)
if args.shownodesbw:
@@ -4197,7 +4218,7 @@ if __name__ == '__main__':
devCsv = formatCsv(deviceList)
print(devCsv)
if not isConciseInfoRequested(args):
if not isConciseInfoRequested(args) and args.showhw == False:
printLogSpacer(footerString)
rsmi_ret_ok(rocmsmi.rsmi_shut_down())