why numpy on new machine so slow #24022
Description
Steps to reproduce:
test_numpy.py
import numpy as np
def testfunc(x):
np.random.seed(x)
X = np.random.randn(5000, 5000)
np.linalg.eigh(X @ X.T)
# %timeit testfunc(0)
import time
for i in range(2):
t1 = time.time()
testfunc(0)
t2 = time.time()
print(t2 - t1)
# break
python test_numpy.py
Error message:
15.402074337005615
15.52828049659729
each take about 15 sec each testfunc call.
soo slow.
1. I have test this simple task on another machine too. (same hardware)
machine01 = 15 sec
machine02 = 1.5sec (10 times faster!!)
2. I have test another simple pure c++ program on these two machinws.
machine01 = 20 sec
machine02 = 21 sec
which means these two machine's computing power is close
so I doubt that, some basic lib is incorrect on machine 01, or something wrong happend.
so far, I have try many solutions but failed.
as bellow Additional information: I have list all useful info for you. Thanks
(what should I check next?, please help!)Additional information:
htop
python --version
python -m cProfile test_numpy.py # check time cost
pip show and config.show()
ldd info
system info
cpus
free -h
