>after 400b cycles to get a simics checkpoint called phase_00*
If you received the simics checkpoints from epfl site, then phase_000 is in your hand. What you did here is called phase_001 >And my command is : There is no need to create flexpoints for this analysis. You can start with the first phase and a trace simulator to run for some million cycles. >I used the sum measurement to get it statics You can run "stat-manager print all" to see the aggregate numbers of cores or "stat-manager -per-node print all" to see the stats per core. On 07/03/2013 19:42, zhangwuxiang wrote: > Hi! > I want to test the cache miss rate of the cloudsuite. And I used flexus simulator. > parameters below: > 0-9 phases , after 400b cycles to get a simics checkpoint called phase_00* > use the trace option "rungen trace" for each phase. > > flexus.set "-magic-break:stop_cycle" "200000000" > flexus.set-region-interval "100000000" > > And my command is : > [root@localhost flexus-4.1]# run_job -postprocess "/home/zhangwuxiang/flexus-4.1/scripts/postprocess_ckptgen.sh trace cassandra" -cfg trace -run trace -local -ckpt-gen CMP.L2Shared.Trace cassandra/2cpu > > I do not know this measure method is right or not. It run 200m cycles with CMP.L2Shared.Trace simulator for each phase. > I used the sum measurement to get it statics. It shows the l1i statics are totally different with different number of cores . For example: > 1 cpu for phase0 of cassandra: > > Nodes-L1i-Hits:Fetch:Owned 0 > Nodes-L1i-Hits:Fetch:Shared 198112078 > Nodes-L1i-Hits:NAW:Exclusive 0 > Nodes-L1i-Hits:NAW:Migratory 0 > Nodes-L1i-Hits:NAW:Modified 0 > Nodes-L1i-Hits:Read:Exclusive 0 > Nodes-L1i-Hits:Read:Migratory 0 > Nodes-L1i-Hits:Read:Modified 0 > Nodes-L1i-Hits:Read:Owned 0 > Nodes-L1i-Hits:Read:Shared 0 > Nodes-L1i-Hits:Upgrade:Exclusive 0 > Nodes-L1i-Hits:Upgrade:Migratory 0 > Nodes-L1i-Hits:Upgrade:Modified 0 > Nodes-L1i-Hits:Write:Exclusive 0 > Nodes-L1i-Hits:Write:Migratory 0 > Nodes-L1i-Hits:Write:Modified 0 > Nodes-L1i-Misses:Evict:Invalid 0 > Nodes-L1i-Misses:EvictD:Invalid 0 > Nodes-L1i-Misses:EvictD:Owned 0 > Nodes-L1i-Misses:EvictD:Shared 0 > Nodes-L1i-Misses:EvictWritable:Invalid 0 > Nodes-L1i-Misses:EvictWritable:Owned 0 > Nodes-L1i-Misses:EvictWritable:Shared 0 > Nodes-L1i-Misses:Fetch:Invalid 1678686 > ...... > 2 cpu for phase0 of cassandra: > > sys-L1i-Hits:Fetch:Owned 0 > sys-L1i-Hits:Fetch:Shared 994919426 > sys-L1i-Hits:NAW:Exclusive 0 > sys-L1i-Hits:NAW:Migratory 0 > sys-L1i-Hits:NAW:Modified 0 > sys-L1i-Hits:Read:Exclusive 0 > sys-L1i-Hits:Read:Migratory 0 > sys-L1i-Hits:Read:Modified 0 > sys-L1i-Hits:Read:Owned 0 > sys-L1i-Hits:Read:Shared 0 > sys-L1i-Hits:Upgrade:Exclusive 0 > sys-L1i-Hits:Upgrade:Migratory 0 > sys-L1i-Hits:Upgrade:Modified 0 > sys-L1i-Hits:Write:Exclusive 0 > sys-L1i-Hits:Write:Migratory 0 > sys-L1i-Hits:Write:Modified 0 > sys-L1i-Misses:Evict:Invalid 0 > sys-L1i-Misses:EvictD:Invalid 0 > sys-L1i-Misses:EvictD:Owned 0 > sys-L1i-Misses:EvictD:Shared 0 > sys-L1i-Misses:EvictWritable:Invalid 0 > sys-L1i-Misses:EvictWritable:Owned 0 > sys-L1i-Misses:EvictWritable:Shared 0 > sys-L1i-Misses:Fetch:Invalid 4896786 > .... > Why are they so different? Is anything wrong with my measure method? Please let me know how to test the cache miss rate. Thank you. 2013-07-03 > ------------------------- > > Best Regards! > MAIL: [email protected] [1] -- Regards, Mahmood Links: ------ [1] mailto:[email protected]
