DEPRECATED: Please see https://github.com/IntelPython/mkl-service for the current project.

This package exposes a few functions which are declared in mkl_service.h. The main purpose of the package is to allow the user to change the number of CPU's MKL is using at runtime.

In this small example, we first query how many CPU are being used, and then change this number.

>>> import mkl
>>> mkl.get_max_threads()
4
>>> mkl.set_num_threads(1)
>>> mkl.get_max_threads()
1

mkl service functions:

mkl.get_cpu_clocks() -> int

Return the CPU clocks as an integer.

mkl.get_cpu_frequency() -> float

Return CPU frequency in GHz as a float.

mkl.get_max_threads() -> int

Return the number of threads Intel MKL is targeting for parallelism.

mkl.get_version_string() -> str

Return the MKL library version information as a string.

mkl.mem_stat() -> (int, int)

Returns a tuple (bytes, count) containing memory usage statistics of the MKL allocator - number of bytes allocated (bytes). - number of allocated blocks (count).

mkl.set_num_threads(n)

Set the number of threads MKL should use. This is only a hint, and no guaranteed is made this number of threads will actually be used. This function takes precedence over the environment variable MKL_NUM_THREADS.