Since all communication is attained through the copy and reduce operations, their implementation is clearly critical to the performance of higher level PLAPACK routines, to be covered in subsequent chapters, as well as applications built on the PLAPACK infrastructure. The number of special cases that must be accommodated by these routines is considerable. For this reason, we will show how the copy and reduce can be implemented in a very systematic way, which greatly reduces complexity, while providing reasonable performance. Next we will show how high performance can be attained for certain frequently encountered cases, by introducing optimized implementations.