Incollection,
Database Scan Variants on Modern CPUs: A Performance Study
In Memory Data Management and Analysis, Springer International Publishing, (2015)
Abstract
Main-memory databases rely on highly tuned database operations
to achieve peak performance. Recently, it has been shown that
different code optimizations for database operations favor
different processors. However, it is still not clear how the
combination of code optimizations (e.g., loop unrolling and
vectorization) will affect the performance of database
algorithms on different processors. In this paper, we extend
prior studies by an in-depth performance analysis of different
variants of the scan operator. We find that the performance of
the scan operator for different processors gets even harder to
predict when multiple code optimizations are combined. Since the
scan is the most simple database operator, we expect the same
effects for more complex operators such as joins. Based on these
results, we identify practical problems for a query processor
and discuss how we can counter these challenges in future work.
