Hi, I have several questions with regarding to the cache warmup option in Gem5.
My command to run the benchmark with cache warmup is: $GEM5/build/X86/gem5.opt -d Output/m5out_test $GEM5/configs/example/se.py --cpu-type \"detailed\" --caches --l2cache -n 1 --bench "omnetpp" -F 2000000000 -standard-switch=100000000 -warmup-insts 100000000 -I 100000000 --l2_size='1MB' --l2_assoc=4 I changed the number of warmup instructions for different simulations. However, I always keep the number of instructions of (fastforward+warmup) as 2.1 billion so that the real simulations have the same start point. I observed the following facts from my results. 1. There is a big difference (as much as 64%) in numCycles between the simulations w/ cache warmup and w/o cache warmup. 2. No matter how I changed the number of warmup instructions, the numCycles results did not change. I tried 100M, 50M, 10M, 5M, 1M, 1000, and even 20 for the number of warmup instructions. Thus, I'm a little suspicious about the experiment results and I have the following questions: 1. What does the number following "--standard-switch" mean? How should I configure it? I tried to change it, but it seems that this number doesn't change the simulation results. 2. Is there any misuse in my experiments about cache warmup? Are the results reasonable? Thanks! Any help would be appreciated! Sean
_______________________________________________ gem5-users mailing list [email protected] http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users
