Hi all, I was looking at my simulation results (default configuration without modifying marss) to extract the L1 data cache access frequency in SPEC2006 benchmarks. But I saw some strange value in the output yaml file. In the following I show the extracted data using mstats.py for astar benchmark:
astar.user:base_machine:ooo_0_0:thread0:commit:opclass:ld:17093680 astar.user:base_machine:L1_D_0:cpurequest:count:hit:read:forward:0 astar.user:base_machine:L1_D_0:cpurequest:count:hit:read:hit:18322669 astar.user:base_machine:L1_D_0:cpurequest:count:miss:read:76062 18322669 + 76062 = 18398731, 7.6% more than committed loads due to speculation (my guess) astar.user:base_machine:ooo_0_0:thread0:commit:opclass:st:10130727 astar.user:base_machine:L1_D_0:cpurequest:count:hit:write:forward:0 astar.user:base_machine:L1_D_0:cpurequest:count:hit:write:hit:2491218 astar.user:base_machine:L1_D_0:cpurequest:count:miss:write:135879 2491218 + 135879 = 2627097, -74% less than committed stores!!! I have seen similar behaviors in other benchmarks. I am not sure if I am looking at correct values. I greatly appreciate any help or explanation. Thanks, Mostafa
_______________________________________________ http://www.marss86.org Marss86-Devel mailing list [email protected] https://www.cs.binghamton.edu/mailman/listinfo/marss86-devel
