Wed 17 Jun 2015 14:50 - 15:15 at PLDI Main BLUE (Portland 254-255) - Parallelism Chair(s): Sara Baghsorkhi

This paper introduces two new compiler transformations for representing and transforming sparse matrix computations and their data representations. In cooperation with run-time inspection, our compiler derives new sparse matrix representations and associated transformed code to implement a variety of representations targeting different architecture platforms. This systematic approach to combining code and data transformations on sparse computations, which extends a polyhedral transformation and code generation framework, permits the compiler to compose these transformations with other transformations to generate high-performing code that is within 5\% and often exceeds manually-tuned, high-performance sparse matrix libraries CUSP and OSKI. Additionally, our compiler-generated inspector codes are on average 1.5x faster than OSKI and match that of CUSP respectively.