Speed up `SimpleArray::matmul()`

[yungyuc]

The current implementation of SimpleArray::matmul() uses the naive algorithm, see issue #587. A benchmark to vendor library and profiling should be performed to develop an optimized implementation.

This issue follows up issue #588.

Reference

• Apple accelerate (vendor library): https://developer.apple.com/documentation/accelerate

[ThreeMonth03]

Hi, @yungyuc . I'm interested in this issue.

[yungyuc]

No problem. When do you plan to make some progress?

[ThreeMonth03]

No problem. When do you plan to make some progress.

I will survey and profile the benchmark before this Saturday. As for the comparision, I will profile the matrix multiplication operation of 1-dimensional and 2-dimensional array.

[yungyuc]

Please focus on matrix-matrix (2D array) multiplication. Matrix-vector multiplication is not interesting.