I use the SPEC CPU2006 benchmarks for evaluation in my graduation project. My upperclassman asked me to sort out a table consisting of MPKI, RBL and BLP of SPEC CPU2006 benchmarks.
This is an explanation of the above abbreviated words:
- MPKI: Misses per kiloinstruction
- RBL: Row-buffer locality
- BLP: Bank-level parallelism
I referred to some papers to obtain these data.
Here is the table:
| Benchmark | MPKI | RBL | BLP |
|---|---|---|---|
| 429.mcf | 97.38% | 42.41 | 6.2 |
| 462.libquantum | 50.00% | 99.22 | 1.05 |
| 437.leslie3d | 49.35% | 91.18 | 1.51 |
| 450.soplex | 46.70% | 88.84 | 1.79 |
| 470.lbm | 43.52% | 95.17 | 2.82 |
| 459.GemsFDTD | 31.79% | 56.22 | 3.15 |
| 482.sphinx3 | 24.94% | 84.78 | 2.24 |
| 483.xalancbmk | 22.95% | 72.01 | 2.35 |
| 471.omnetpp | 21.63% | 45.71 | 4.37 |
| 436.cactusADM | 12.01% | 19.05 | 1.43 |
| 473.astar | 9.26% | 75.24 | 1.61 |
| 456.hmmer | 5.66% | 34.42 | 1.25 |
| 401.bzip2 | 3.98% | 71.44 | 1.87 |
| 464.h264ref | 2.3 | 90.34% | 1.19 |
| 435.gromacs | 0.98 | 89.25% | 1.54 |
| 445.gobmk | 0.77 | 65.76% | 1.52 |
| 458.sjeng | 0.39 | 12.47% | 1.57 |
| 403.gcc | 0.34 | 70.92% | 1.96 |
| 447.dealII | 0.21 | 86.83% | 1.22 |
| 481.wrf | 0.21 | 92.34% | 1.23 |
| 444.namd | 0.19 | 93.05% | 1.16 |
| 400.perlbench | 0.12 | 81.59% | 1.66 |
| 454.calculix | 0.1 | 88.71% | 1.2 |
| 465.tonto | 0.03 | 88.60% | 1.81 |
| 453.povray | 0.01 | 87.22% | 1.43 |
References:
- Thread Cluster Memory Scheduling Exploiting Differences in Memory Access Behavior
- Fairness via Source Throttling A Configurable and High-Performance Fairness Substrate for Multicore Memory Systems
- Staged Memory Scheduling Achieving High Performance and Scalability in Heterogeneous Systems
