BlackjackBench: portable hardware characterization with automated results analysis. To be published in The

DARPA’s AACE project aimed to develop Architecture Aware Com-piler Environments that automatically characterizes the hardware and optimizes the application codes accordingly. We present the BlackjackBench – a suite of portable benchmarks that automate system characterization, plus statistical analysis techniques for in-terpreting the results. The BlackjackBench discovers the effective sizes and speeds of the hardware environment rather than the of-ten unattainable peak values. We aim at hardware characteristics that can be observed by running standard C codes. We characterize the memory hierarchy, including cache sharing and NUMA char-acteristics of the system, properties of the processing cores affect-ing instruction execution speed, and the length of the OS scheduler time slot. We show how they all could potentially interfere with each other and how established classification and statistical analy-sis techniques reduce experimental noise and aid automatic inter-pretation of results. 1