Performance of the decoupled ACRI-1 architecture: the perfect club

This paper examines the performance potential of decoupled computer architectures on real-world codes, and includes the rst performance bounds calculations to be published for the highly-decoupled ACRI-1 computer architecture. It also constitutes the rst published work to report on the eeectiveness of a decoupling Fortran90 compiler. Decoupling is an architectural optimisation which ooers very high sustained performance through large-scale latency hiding. This paper investigates the applicability of access and control decoupling to real-world codes. We illustrate this with compiler-generated decoupling optimisa-tions for the Perfect Club benchmark suite on the Advanced Computer Research Institute's ACRI-1 system, utilising the frequency of loss of de-coupling (LOD) events as a measure of the eeectiveness of decoupling to each code. We derive bounds for the performance of these codes and show that, whilst some exhibit performance roughly equivalent to that on vector computers, others exhibit considerably higher performance potential in a decoupled system.