topic You are right, you may use in Intel® oneAPI Math Kernel Library

How to implement numpy broadcast mechanism with mkl?

xianfeng__liang — Fri, 04 Jan 2019 14:47:01 GMT

How to implement numpy broadcast mechanism with mkl?

I have been confused, how to use mkl to efficiently implement the broadcast mechanism in numpy ((Element wise operator "+","-","*")?
such as
2-D array sub 1-D array
[[1,2,3],
[4,5,6],
[7,8,9]]
-
[1, 2, 3]
=
[[0, 0, 0],
[3, 3, 3],
[6, 6, 6]]

And the second operation (can be understood as a matrix multiplied by a diagonal matrix)
2-D array multiply 1-D array(Element wise multiply )
[[1,2,3],
[4,5,6],
[7,8,9]]
*
[1, 2, 3]
=
[[1, 4, 9],
[4, 10, 18],
[7, 16, 27]]

I tried to implement with the for loop +cblas_dscal/vdSub
But I think this is not efficient, I don't know if there is any better implementation.

You are right, you may use

Gennady_F_Intel — Mon, 07 Jan 2019 05:28:09 GMT

You are right, you may use loop +cblas_dscal/vdSub but this approach will not be efficient. current version of MKL doesn't provide such functionality,

in the case if user

Gennady_F_Intel — Mon, 07 Jan 2019 05:32:00 GMT

in the case if user experiences some performance problem when using numpy, scipy or etc -- Our recommendation to try optimized version of Python where many of math operations are optimized by Intel Performance libraries (IPP, MKL and DAAL) as a backend.

https://software.intel.com/en-us/distribution-for-python