The Alberta Workloads for the SPEC CPU2017 Benchmark Suite

SPEC CPU2017 Benchmarks

Benchmarks With Full Reports
Benchmark NameFindings SummaryKit UsedLink
502.gcc_r That the benchmark behaviour does not change significantly with different workloads. 93 gcc report
505.mcf_r The benchmark shows different behaviour using new inputs. 91 mcf report
507.cactuBSSN_r Most workloads are fairly similar, excluding changes to one parameter. 93 cactuBSSN report
510.parest_r Banchmark's behaviour may depend on some parameters more than others.. 93 parest report
511.povray_r Benchmark's behavior is very different from workload to workload, while the train and refrate workloads are nearly identical. 93 povray report
519.lbm_r Behavior of the benchmark does not vary significantly when workload changes. 93 lbm report
520.omnetpp_r Behavior of the benchmark does not vary significantly between workloads. 93 omnetpp report
523.xalancbmk_r Execution is very different from workload to workload. 93 xalancbmk report
525.x264_r Execution is similar between workloads. 93 x264 report
531.deepsjeng_r Found that the benchmark's behavior does not vary greatly between workloads. 93 deepsjeng report
541.leela_r Found that, other than memory accesses, the behavior of the benchmark varies little between workloads. 93 leela report
544.nab_r Found that the benchmark's behavior does not vary greatly between workloads. 93 nab report
548.exchange2_r Found that the benchmark's behavior does not vary greatly from workload to workload. 93 exchange2 report
557.xz_r Found that the number of calls to the benchmark's main functions can vary greatly between workloads. 93 xz report
Benchmarks With Partial Reports
Benchmark NameFindings SummaryUnable to CompileKit UsedLink
521.wrf_r Execution varies more when parameters are changed than with changes to input dataset. ICC 93 wrf Report
526.blender_r Found that the train workload is extremely short, and that the blender benchmark will behave extremely different depending on the input files. GCC and LLVM 74 blender Report
Benchmarks With Reported Issues
Benchmark NameFindings SummaryAddressedLink
500.perlbench_r Most significant Perl applications available include substantial C code. Given that the benchmark is not designed to handle C code, such applications are not suitable as alternative workloads for the benchmark. Thus we were unable to provide new inputs for this benchmark. No Insights Report