Matrix multiplication is usually done by a platform-specific BLAS library (BLAS is an API, there are multiple implementations, e.g. Intel MKL, OpenBLAS, cuBLAS)

Matrix multiplication is usually done by a platform-specific BLAS library (BLAS is an API, there are multiple implementations, e.g. Intel MKL, OpenBLAS, cuBLAS). There are some other linear algebra APIs/libraries, but this is what's used the most. Most of the numerical code that cares about performance for linear algebra uses this API and links an appropriate implementation.