Merge amd-staging into amd-master 20240508
Signed-off-by: Maisam Arif <Maisam.Arif@amd.com> Change-Id: I8ce1757eb0666c7c3242556e20c0fd5de22d740b
Этот коммит содержится в:
Maisam Arif
@@ -2,8 +2,9 @@
|
||||
|
||||
This tool acts as a command line interface for manipulating
|
||||
and monitoring the amdgpu kernel, and is intended to replace
|
||||
and deprecate the existing rocm_smi.py CLI tool.
|
||||
It uses Ctypes to call the rocm_smi_lib API.
|
||||
and deprecate the existing rocm_smi.py CLI tool located at
|
||||
https://github.com/ROCm/ROC-smi.
|
||||
This tool uses Ctypes to call the rocm_smi_lib API.
|
||||
Recommended: At least one AMD GPU with ROCm driver installed
|
||||
Required: ROCm SMI library installed (librocm_smi64)
|
||||
|
||||
|
||||
@@ -3,8 +3,9 @@
|
||||
|
||||
This tool acts as a command line interface for manipulating
|
||||
and monitoring the amdgpu kernel, and is intended to replace
|
||||
and deprecate the existing rocm_smi.py CLI tool.
|
||||
It uses Ctypes to call the rocm_smi_lib API.
|
||||
and deprecate the existing rocm_smi.py CLI tool located at
|
||||
https://github.com/ROCm/ROC-smi.
|
||||
This tool uses Ctypes to call the rocm_smi_lib API.
|
||||
Recommended: At least one AMD GPU with ROCm driver installed
|
||||
Required: ROCm SMI library installed (librocm_smi64)
|
||||
"""
|
||||
@@ -31,7 +32,7 @@ from rsmiBindings import *
|
||||
# Patch version - Increment when adding a fix, set to 0 when minor is incremented
|
||||
# Hash version - Shortened commit hash. Print here and not with lib for consistency with amd-smi
|
||||
SMI_MAJ = 2
|
||||
SMI_MIN = 1
|
||||
SMI_MIN = 2
|
||||
SMI_PAT = 0
|
||||
# SMI_HASH is provided by rsmiBindings
|
||||
__version__ = '%s.%s.%s+%s' % (SMI_MAJ, SMI_MIN, SMI_PAT, SMI_HASH)
|
||||
@@ -1269,6 +1270,34 @@ def setClockExtremum(deviceList, level, clkType, clkValue, autoRespond):
|
||||
printLog(device, 'Setting %s %s clock is not supported for this device.' % (level, clkType), None)
|
||||
|
||||
|
||||
def setVoltageCurve(deviceList, point, clk, volt, autoRespond):
|
||||
""" Set voltage curve for a point in the PowerPlay table for a list of devices.
|
||||
|
||||
:param deviceList: List of DRM devices (can be a single-item list)
|
||||
:param point: Point on the voltage curve to modify
|
||||
:param clk: Clock speed specified for this curve point
|
||||
:param volt: Voltage specified for this curve point
|
||||
:param autoRespond: Response to automatically provide for all prompts
|
||||
"""
|
||||
global RETCODE
|
||||
value = '%s %s %s' % (point, clk, volt)
|
||||
try:
|
||||
any(int(item) for item in value.split())
|
||||
except ValueError:
|
||||
printErrLog(None, 'Unable to set Voltage curve')
|
||||
printErrLog(None, 'Non-integer characters are present in %s' %value)
|
||||
RETCODE = 1
|
||||
return
|
||||
confirmOutOfSpecWarning(autoRespond)
|
||||
for device in deviceList:
|
||||
ret = rocmsmi.rsmi_dev_od_volt_info_set(device, int(point), int(clk), int(volt))
|
||||
if rsmi_ret_ok(ret, device, 'set_voltage_curve'):
|
||||
printLog(device, 'Successfully set voltage point %s to %s(MHz) %s(mV)' % (point, clk, volt), None)
|
||||
else:
|
||||
printErrLog(device, 'Unable to set voltage point %s to %s(MHz) %s(mV)' % (point, clk, volt))
|
||||
RETCODE = 1
|
||||
|
||||
|
||||
def setPowerPlayTableLevel(deviceList, clkType, point, clk, volt, autoRespond):
|
||||
""" Set clock frequency and voltage for a level in the PowerPlay table for a list of devices.
|
||||
|
||||
@@ -2682,6 +2711,38 @@ def showPower(deviceList):
|
||||
printLogSpacer()
|
||||
|
||||
|
||||
def showPowerPlayTable(deviceList):
|
||||
""" Display current GPU Memory clock frequencies and voltages for a list of devices
|
||||
|
||||
:param deviceList: List of DRM devices (can be a single-item list)
|
||||
"""
|
||||
global PRINT_JSON
|
||||
if PRINT_JSON:
|
||||
return
|
||||
printLogSpacer(' GPU Memory clock frequencies and voltages ')
|
||||
odvf = rsmi_od_volt_freq_data_t()
|
||||
for device in deviceList:
|
||||
ret = rocmsmi.rsmi_dev_od_volt_info_get(device, byref(odvf))
|
||||
if rsmi_ret_ok(ret, device, 'get_od_volt'):
|
||||
# TODO: Make this more dynamic and less hard-coded if possible
|
||||
printLog(device, 'OD_SCLK:', None)
|
||||
printLog(device, '0: %sMhz' % (int(odvf.curr_sclk_range.lower_bound / 1000000)), None)
|
||||
printLog(device, '1: %sMhz' % (int(odvf.curr_sclk_range.upper_bound / 1000000)), None)
|
||||
printLog(device, 'OD_MCLK:', None)
|
||||
printLog(device, '0: %sMhz' % (int(odvf.curr_mclk_range.lower_bound / 1000000)), None)
|
||||
printLog(device, '1: %sMhz' % (int(odvf.curr_mclk_range.upper_bound / 1000000)), None)
|
||||
if odvf.sclk_freq_limits.lower_bound > 0 or odvf.sclk_freq_limits.upper_bound > 0 \
|
||||
or odvf.mclk_freq_limits.lower_bound >0 or odvf.mclk_freq_limits.upper_bound > 0:
|
||||
printLog(device, 'OD_RANGE:', None)
|
||||
if odvf.sclk_freq_limits.lower_bound > 0 or odvf.sclk_freq_limits.upper_bound > 0:
|
||||
printLog(device, 'SCLK: %sMhz %sMhz' % (
|
||||
int(odvf.sclk_freq_limits.lower_bound / 1000000), int(odvf.sclk_freq_limits.upper_bound / 1000000)), None)
|
||||
if odvf.mclk_freq_limits.lower_bound >0 or odvf.mclk_freq_limits.upper_bound > 0:
|
||||
printLog(device, 'MCLK: %sMhz %sMhz' % (
|
||||
int(odvf.mclk_freq_limits.lower_bound / 1000000), int(odvf.mclk_freq_limits.upper_bound / 1000000)), None)
|
||||
printLogSpacer()
|
||||
|
||||
|
||||
def showProduct(deviceList):
|
||||
""" Show the requested product information for a list of devices
|
||||
|
||||
@@ -2753,7 +2814,7 @@ def showRange(deviceList, rangeType):
|
||||
:param rangeType: [sclk|voltage] Type of range to return
|
||||
"""
|
||||
global RETCODE
|
||||
if rangeType not in {'sclk', 'mclk'}:
|
||||
if rangeType not in {'sclk', 'mclk', 'voltage'}:
|
||||
printLog(None, 'Invalid range identifier %s' % (rangeType), None)
|
||||
RETCODE = 1
|
||||
return
|
||||
@@ -2768,6 +2829,21 @@ def showRange(deviceList, rangeType):
|
||||
if rangeType == 'mclk':
|
||||
printLog(device, 'Valid mclk range: %sMhz - %sMhz' % (
|
||||
int(odvf.curr_mclk_range.lower_bound / 1000000), int(odvf.curr_mclk_range.upper_bound / 1000000)), None)
|
||||
if rangeType == 'voltage':
|
||||
if odvf.num_regions == 0:
|
||||
printErrLog(device, 'Voltage curve regions unsupported.')
|
||||
continue
|
||||
num_regions = c_uint32(odvf.num_regions)
|
||||
regions = (rsmi_freq_volt_region_t * odvf.num_regions)()
|
||||
ret = rocmsmi.rsmi_dev_od_volt_curve_regions_get(device, byref(num_regions), byref(regions))
|
||||
if rsmi_ret_ok(ret, device, 'volt'):
|
||||
for i in range(num_regions.value):
|
||||
printLog(device,
|
||||
'Region %d: Valid voltage range: %smV - %smV' % (i, regions[i].volt_range.lower_bound,
|
||||
regions[i].volt_range.upper_bound),
|
||||
None)
|
||||
else:
|
||||
printLog(device, 'Unable to display %s range' % (rangeType), None)
|
||||
printLogSpacer()
|
||||
|
||||
|
||||
@@ -3085,6 +3161,25 @@ def showVoltage(deviceList):
|
||||
printLogSpacer()
|
||||
|
||||
|
||||
def showVoltageCurve(deviceList):
|
||||
""" Show the voltage curve points for the specified devices
|
||||
|
||||
:param deviceList: List of DRM devices (can be a single-item list)
|
||||
"""
|
||||
printLogSpacer(' Voltage Curve Points ')
|
||||
odvf = rsmi_od_volt_freq_data_t()
|
||||
for device in deviceList:
|
||||
ret = rocmsmi.rsmi_dev_od_volt_info_get(device, byref(odvf))
|
||||
if rsmi_ret_ok(ret, device, 'get_od_volt_info', silent=False) and odvf.num_regions > 0:
|
||||
for position in range(3):
|
||||
printLog(device, 'Voltage point %d: %sMhz %smV' % (
|
||||
position, int(list(odvf.curve.vc_points)[position].frequency / 1000000),
|
||||
int(list(odvf.curve.vc_points)[position].voltage)), None)
|
||||
else:
|
||||
printErrLog(device, 'Voltage curve Points unsupported.')
|
||||
printLogSpacer()
|
||||
|
||||
|
||||
def showXgmiErr(deviceList):
|
||||
""" Display the XGMI Error status
|
||||
|
||||
@@ -3738,6 +3833,7 @@ if __name__ == '__main__':
|
||||
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')
|
||||
groupDisplayTop.add_argument('--showbus', help='Show PCI bus number', action='store_true')
|
||||
groupDisplayPages.add_argument('--showpagesinfo', help='Show retired, pending and unreservable pages',
|
||||
action='store_true')
|
||||
@@ -3762,6 +3858,8 @@ if __name__ == '__main__':
|
||||
groupDisplay.add_argument('-o', '--showoverdrive', help='Show current GPU Clock OverDrive level',
|
||||
action='store_true')
|
||||
groupDisplay.add_argument('-p', '--showperflevel', help='Show current DPM Performance Level', action='store_true')
|
||||
groupDisplay.add_argument('-S', '--showclkvolt', help='Show supported GPU and Memory Clocks and Voltages',
|
||||
action='store_true')
|
||||
groupDisplay.add_argument('-s', '--showclkfrq', help='Show supported GPU and Memory Clock', action='store_true')
|
||||
groupDisplay.add_argument('--showmeminfo', help='Show Memory usage information for given block(s) TYPE',
|
||||
metavar='TYPE', type=str, nargs='+')
|
||||
@@ -3773,6 +3871,7 @@ if __name__ == '__main__':
|
||||
groupDisplay.add_argument('--showrasinfo',
|
||||
help='Show RAS enablement information and error counts for the specified block(s) (all if no arg given)',
|
||||
nargs='*')
|
||||
groupDisplay.add_argument('--showvc', help='Show voltage curve', action='store_true')
|
||||
groupDisplay.add_argument('--showxgmierr', help='Show XGMI error information since last read', action='store_true')
|
||||
groupDisplay.add_argument('--showtopo', help='Show hardware topology information', action='store_true')
|
||||
groupDisplay.add_argument('--showtopoaccess', help='Shows the link accessibility between GPUs ', action='store_true')
|
||||
@@ -3812,6 +3911,8 @@ if __name__ == '__main__':
|
||||
groupAction.add_argument('--setmlevel',
|
||||
help='Change GPU Memory clock frequency (MHz) and Voltage for (mV) a specific Level',
|
||||
metavar=('MCLKLEVEL', 'MCLK', 'MVOLT'), nargs=3)
|
||||
groupAction.add_argument('--setvc', help='Change SCLK Voltage Curve (MHz mV) for a specific point',
|
||||
metavar=('POINT', 'SCLK', 'SVOLT'), nargs=3)
|
||||
groupAction.add_argument('--setsrange', help='Set min and max SCLK speed', metavar=('SCLKMIN', 'SCLKMAX'), nargs=2)
|
||||
groupAction.add_argument('--setextremum', help='Set min/max of SCLK/MCLK speed', metavar=('min|max', "sclk|mclk", 'CLK'), nargs=3)
|
||||
groupAction.add_argument('--setmrange', help='Set min and max MCLK speed', metavar=('MCLKMIN', 'MCLKMAX'), nargs=2)
|
||||
@@ -3881,7 +3982,7 @@ if __name__ == '__main__':
|
||||
or args.resetclocks or args.setprofile or args.resetprofile or args.setoverdrive or args.setmemoverdrive \
|
||||
or args.setpoweroverdrive or args.resetpoweroverdrive or args.rasenable or args.rasdisable or \
|
||||
args.rasinject or args.gpureset or args.setperfdeterminism or args.setslevel or args.setmlevel or \
|
||||
args.setsrange or args.setextremum or args.setmrange or args.setclock or \
|
||||
args.setvc or args.setsrange or args.setextremum or args.setmrange or args.setclock or \
|
||||
args.setcomputepartition or args.setmemorypartition or args.resetcomputepartition or args.resetmemorypartition:
|
||||
relaunchAsSudo()
|
||||
|
||||
@@ -3928,6 +4029,7 @@ if __name__ == '__main__':
|
||||
args.showproductname = True
|
||||
args.showserial = True
|
||||
args.showuniqueid = True
|
||||
args.showvoltagerange = True
|
||||
args.showbus = True
|
||||
args.showpagesinfo = True
|
||||
args.showfan = True
|
||||
@@ -3945,12 +4047,14 @@ if __name__ == '__main__':
|
||||
args.showpids = "summary"
|
||||
args.showpidgpus = []
|
||||
args.showreplaycount = True
|
||||
args.showvc = True
|
||||
args.showcomputepartition = True
|
||||
args.showmemorypartition = True
|
||||
|
||||
if not PRINT_JSON:
|
||||
args.showprofile = True
|
||||
args.showclkfrq = True
|
||||
args.showclkvolt = True
|
||||
|
||||
# Don't do reset in combination with any other command
|
||||
if args.gpureset:
|
||||
@@ -4021,6 +4125,8 @@ if __name__ == '__main__':
|
||||
showPids(args.showpids)
|
||||
if args.showpidgpus or str(args.showpidgpus) == '[]':
|
||||
showGpusByPid(args.showpidgpus)
|
||||
if args.showclkvolt:
|
||||
showPowerPlayTable(deviceList)
|
||||
if args.showvoltage:
|
||||
showVoltage(deviceList)
|
||||
if args.showbus:
|
||||
@@ -4064,6 +4170,10 @@ if __name__ == '__main__':
|
||||
showRange(deviceList, 'sclk')
|
||||
if args.showmclkrange:
|
||||
showRange(deviceList, 'mclk')
|
||||
if args.showvoltagerange:
|
||||
showRange(deviceList, 'voltage')
|
||||
if args.showvc:
|
||||
showVoltageCurve(deviceList)
|
||||
if args.showenergycounter:
|
||||
showEnergy(deviceList)
|
||||
if args.showcomputepartition:
|
||||
@@ -4100,6 +4210,8 @@ if __name__ == '__main__':
|
||||
resetPowerOverDrive(deviceList, args.autorespond)
|
||||
if args.setprofile:
|
||||
setProfile(deviceList, args.setprofile)
|
||||
if args.setvc:
|
||||
setVoltageCurve(deviceList, args.setvc[0], args.setvc[1], args.setvc[2], args.autorespond)
|
||||
if args.setextremum:
|
||||
setClockExtremum(deviceList, args.setextremum[0], args.setextremum[1], args.setextremum[2], args.autorespond)
|
||||
if args.setsrange:
|
||||
|
||||
Ссылка в новой задаче
Block a user