The characterization of two scientific workloads using the CRAY X-MP performance monitor

The week-end production period on a Cray X-MP was monitored for several months at each of two supercomputing sites. The hardware performance monitor available on the X-MP was used to collect the data at each site. Various metrics are computed using the measured data. These metrics include rates, such as mops mflops, percent vectorization, average vector lengths, memory reference rates, memory contention, the operation mix, I/O rates, hardware metrics such as operations per clock period, and percentage of time waiting to issue an instruction that is waiting on some hardware component. These metrics are used in evaluating the performance of the workloads and the X-MP architecture. These metrics may also give insight into the design of similar register-based architectures and may be useful in parameterizing models in future work