I’ve seen on a lot of articles about BLAS libraries/implementations that they provide a “kernel” or “computational kernel” (sometimes) :

• https://en.wikipedia.org/wiki/Basic_Linear_Algebra_Subprograms#Background
• Articles such as this one : https://www.sciencedirect.com/science/article/abs/pii/016781919400030E
• Maybe the ‘K’ in the name “Intel MKL” ?

But I have not found a description of what would be a “kernel”. I do not understand the difference with a BLAS library or an implementation.

I know :

• the kernel in linear algebra (null space), is this related ?
• the kernel like in convolutions : https://en.wikipedia.org/wiki/Kernel_(image_processing)
• the kernel in statistics : https://en.wikipedia.org/wiki/Kernel_(statistics)