SIESTA was born in 1996 to implement, in self consistent DFT, the order-N techniques developed for tight-binding in the early 1990s. Because order-N was carefully imposed in all the code, this was very efficient, specially for large systems and moderate precisions. Later, the need for higher precisions, and the addition of many new functionalities, imposed compromises to the order-N philosophy, but efficiency kept being one of SIESTA's strongest points.
Initially, SIESTA was a purely serial code. This was convenient for many "modest" users, who used single workstations, and this gave it a rapid popularity. Later, parallelism was added "on top", with priority given to not compromising serial execution, and without revising the algorithms for good parallelism. Today, parallel computers and clusters have become ubiquitous, and a new parallelization of SIESTA has become a priority. In this effort, the Barcelona Supercomputing Center (BSC) has become a key partner. SIESTA is now an integral part of the BSC software stack and users can apply for BSC projects to run SIESTA.
http://www.icmab.es/siesta/siesta_project
