Because DVMC primarily targets high-availability
commercial servers, we chose the Wisconsin Commercial Workload Suite [2] for our benchmarks. These
workloads are described briefly in Table 7 and in more
detail by Alameldeen et al. [2]. Although SPARC v9 is a
64-bit architecture, portions of code in the benchmark
suite were written for the 32-bit SPARC v8 instruction
set. Since these code segments were written for TSO, a
system configured for PSO or RMO must switch to TSO
while executing 32-bit code. Table 7 shows the average
fraction of 32-bit memory operations executed for each
benchmark during our experiments.
To handle the runtime variability inherent in commercial workloads, we run each simulation ten times
with small pseudo-random perturbations. Our experimental results show mean result values as well as error
bars that correspond to one standard deviation.