Hello, I am currently working on a project that aims at analyzing the differences in SAT/SMT solver performance on benchmarks (IPC/cache misses/branch misses) when varying architectural aspects of a CPU (cache size, associativity, etc). I figured that to begin testing, I would implement a baseline processor in gem5, run in SE mode, and acquire some initial data from the performance files that gem5 generates after a run (i.e. from stats.txt). It is important to note that this project is for a class, so I am searching for the fastest way to gather data, as I only have a few week’s time.
When running a certain benchmark on the glucose binary (with the command ./glucose <benchmark.cnf>), my local machine determines that the benchmark is unsatisfiable within 25 seconds, deterministically. The following line is what I used to run the Glucose SAT solver on gem5: build/X86/gem5.fast configs/example/se.py --cpu-type=DerivO3CPU --cpu-clock=3.4GHz --sys-clock=2GHz --caches --l1d_size=32kB --l1i_size=32kB --l1i_assoc=8 --l1d_assoc=8 --l2cache --l2_size=256kB --l2_assoc=8 --mem-type=DDR4_2400_8x8 --mem-size=16GB -c ./glucose -o 0dc468413b65339fad7e4f981bf2fb1e-reconf20_61_myciel4_2-sat30s.cnf When running the preceding line, I attained output that matched my local run, but it took more than 3 days to finish. It appeared to produce correct output (I ended up killing the run when it was close to finishing, but it had identical run characteristics to the local run).* I was wondering whether this seems like a normal amount of time for a binary SAT program to solve a benchmark on a gem5. Is there anything I can do to make my run produce results faster?* The following PNG shows the SAT output after a 25second local run on my AMD Ryzen7 5800X Processor (LEFT) vs SAT output on gem5 (RIGHT) using the command mentioned above, running for *~20 hours*: Thank you in advance, *--* *Johnny Vallespir* B.S.E. Electrical Engineering - Class of 2022 M.S.E. IC's and VLSI Design - Class of 2023 University of Michigan
_______________________________________________ gem5-users mailing list -- gem5-users@gem5.org To unsubscribe send an email to gem5-users-le...@gem5.org
