mkl_random crash

[garfi303]

I just downloaded Intel python distribution version 2020 Release 1 and have discovered a crashing bug in mkl_random.
Here is the example:

$ bin/python3.7
Python 3.7.7 (default, Mar 13 2020, 13:32:22)
[GCC 7.3.0] :: Intel(R) Corporation on linux
Type "help", "copyright", "credits" or "license" for more information.
Intel(R) Distribution for Python is brought to you by Intel Corporation.
Please check out: https://software.intel.com/en-us/python-distibution
>>> import sys
>>> import mkl_random
>>> sys.version
'3.7.7 (default, Mar 13 2020, 13:32:22) \n[GCC 7.3.0]'
>>> sys.executable
'/mnt/home/hadoop/intelpython3/bin/python3.7'
>>> mkl_random.geometric(1,10)

Intel MKL ERROR: Parameter 5 was incorrect on entry to viRngGeometric.
python3.7: mkl_random/src/mkl_distributions.cpp:1181: void irk_geometric_vec(irk_state_ *, long, int *, double): Assertion `err == 0' failed.
Aborted
$

Compare that with the non-mkl_random:

$ bin/python3.7
Python 3.7.7 (default, Mar 13 2020, 13:32:22)
[GCC 7.3.0] :: Intel(R) Corporation on linux
Type "help", "copyright", "credits" or "license" for more information.
Intel(R) Distribution for Python is brought to you by Intel Corporation.
Please check out: https://software.intel.com/en-us/python-distibution
>>> import numpy
>>> numpy.random.geometric(1,10)
array([1, 1, 1, 1, 1, 1, 1, 1, 1, 1])
>>>

[oleksandr-pavlyk]

Thanks for reporting this.

The fix is committed to the trunk. Please note that that MKL samples are 0-based (non-negative 32-bit integers), while NumPy RNG's are 1-based (positive 64-bit integers).

[oleksandr-pavlyk]

Closing as fixed:

Python 3.7.7 (default, Jun 26 2020, 05:10:03)
Type 'copyright', 'credits' or 'license' for more information
IPython 7.18.1 -- An enhanced Interactive Python. Type '?' for help.

In [1]: import mkl_random
m
In [2]: mkl_random.__version__
Out[2]: '1.2.0'

In [3]: mkl_random.geometric(1, 10)
Out[3]: array([0, 0, 0, 0, 0, 0, 0, 0, 0, 0], dtype=int32)