Creating an optimized environment for data transformations using conda

As you have already seen in Chapter 2, Configuring R with Power BI, the specific operations of numerical linear algebra are handled through the specialized libraries BLAS and LAPACK. These libraries now define a standard interface for those types of operations. But depending on which implementation of them you choose, there can be significant differences in performance.

There are many implementations made under different licenses. For example, the standard BLAS and LAPACK libraries that you often find pre-installed on your system (such as those in CRAN R), do not support multi-threading. The Automatically Tuned Linear Algebra Software (ATLAS) implementation achieves good performance and uses the BSD license. Then there is OpenBLAS, another open-source implementation with very good performance due to multi-threading. Finally, we have the Intel oneAPI Math Kernel Library (MKL), optimized for Intel multi-core...