sascha/workload-automation
1
0
Fork 0
mirror of https://github.com/ARM-software/workload-automation.git synced 2025-10-30 22:54:18 +00:00
Code Issues Releases Wiki Activity

energy_model: yet another adjustment to leakage compensation.

Browse Source
This commit is contained in:
Sergei Trofimov
2015-09-02 15:31:13 +01:00
parent a80780b9ed
commit 3501eccb8e
1 changed files with 15 additions and 28 deletions
Download Patch File Download Diff File Expand all files Collapse all files

43
wlauto/instrumentation/energy_model/__init__.py
View File

@@ -290,7 +290,7 @@ def fit_polynomial(s, n):
return poly(s.index)
def get_cpus_power_table(data, index, opps): # pylint: disable=too-many-locals
def get_cpus_power_table(data, index, opps, leak_factors): # pylint: disable=too-many-locals
# pylint: disable=no-member
power_table = data[[index, 'cluster', 'cpus', 'power']].pivot_table(index=index,
columns=['cluster', 'cpus'],
@@ -307,32 +307,10 @@ def get_cpus_power_table(data, index, opps): # pylint: disable=too-many-locals
bs_power_table.loc[bs_power_table[cluster, 1].notnull(), (cluster, 0)] = \
(2 * power_table[cluster, 1] - power_table[cluster, 2]).values
else:
df = pd.concat([bs_power_table[cluster],
opps[cluster].set_index('frequency')],
axis=1).dropna(subset=[1]).reset_index()
# Create a projection by calculating coefficients from the lowest two OPPs (assume minimal leakage)
v0 = df.voltage[0]
v1 = df.voltage[1]
f0 = df.frequency[0]
f1 = df.frequency[1]
# Assumption:
# P = k1*v*f + k2*v^2*f
coeffs = np.array([
[v0 * f0, (v0**2) * f0],
[v1 * f1, (v1**2) * f1]
])
c1pow = np.array([df[1][0], df[1][1]])
c2pow = np.array([df[2][0], df[2][1]])
c1k1, c1k2 = np.linalg.solve(coeffs, c1pow)
c2k1, c2k2 = np.linalg.solve(coeffs, c2pow)
df['a1'] = pd.Series(df.frequency * df.voltage * c1k1 + df.frequency * df.voltage ** 2 * c1k2,
index=df.index)
df['a2'] = pd.Series(df.frequency * df.voltage * c2k1 + df.frequency * df.voltage ** 2 * c2k2,
index=df.index)
bs_power_table.loc[bs_power_table[cluster, 1].notnull(), (cluster, 0)] = (2 * df.a1 - df.a2).values
leakage = leak_factors[cluster] * 2 * (opps['voltage'] / 1000000)**3 / 0.9**3
leakage_delta = leakage - leakage[0]
bs_power_table.loc[bs_power_table[cluster, 1].notnull(), (cluster, 0)] = \
(2 * power_table[cluster, 1] + leakage_delta - power_table[cluster, 2]).values
# re-order columns and rename colum '0' to 'cluster'
power_table = power_table[sorted(power_table.columns,
@@ -431,6 +409,14 @@ class EnergyModelInstrument(Instrument):
description="""OPP table mapping frequency to volatage (kHz --> mV) for the big cluster."""),
Parameter('little_opps', kind=opp_table,
description="""OPP table mapping frequency to volatage (kHz --> mV) for the little cluster."""),
Parameter('big_leakage', kind=int, default=120,
description="""
Leakage factor for the big cluster (this is specific to a particular core implementation).
"""),
Parameter('little_leakage', kind=int, default=60,
description="""
Leakage factor for the little cluster (this is specific to a particular core implementation).
"""),
]
def validate(self):
@@ -575,7 +561,8 @@ class EnergyModelInstrument(Instrument):
message = 'OPPs not specified for one or both clusters; cluster power will not be adjusted for leakage.'
self.logger.warning(message)
opps = {'big': self.big_opps, 'little': self.little_opps}
measured_cpus_table, cpus_table = get_cpus_power_table(freq_power_table, 'frequency', opps)
leakages = {'big': self.big_leakage, 'little': self.little_leakage}
measured_cpus_table, cpus_table = get_cpus_power_table(freq_power_table, 'frequency', opps, leakages)
measured_cpus_output = os.path.join(output_directory, MEASURED_CPUS_TABLE_FILE)
with open(measured_cpus_output, 'w') as wfh:
measured_cpus_table.to_csv(wfh)