As an aside to this, I was looking at the committed insts stats the other day; the committedInsts stat for the CPU does not count no-nops and instruction pre-fetches. However, the commit stage's commitedInsts stat does. I suppose this makes sense since the CPU's insts stat should probably represent real work done while the commit stage's inst stat should represent anything it's retired. But, I think the comments should make this more clear if that was the intention, if it wasn't the intention I suppose the commit stage's stats should be changed to match the CPU's.
-Tony On Fri, Aug 17, 2012 at 8:03 AM, Ali Saidi <[email protected]> wrote: > committedInstructions are the number of machine instructions that have > been committed. The number of ops is greater than or equal than the number > of instructions. If an instruction is executed without being cracked its an > op, if it is cracked then everything it's cracked into are ops. > sim_ops and sim_insts is a count of all ops instructions that the > simulator has executed (on all CPUs and isn't reset when the statistics > are). It's meant for analyzing the performance of the simulator. > > Ali > > > > On Aug 17, 2012, at 3:32 AM, Ali chaker wrote: > > Hi, > > > I’ve run bbench benchmark using different frequencies (600MHZ, 1.2 GHZ and > 1.5 GHZ) but I’ve low values for IPC and MIPS( MIPS= IPC * Freq). > > What is the difference between *system.cpu.committedInsts *and > *system.cpu.committedOps? > *Why in the IPC Formula we use the CommittedInsts and not the > committedOps? > > What is the difference between (*system.cpu.committedOps* <<< *sim_ops*) > and between (*system.cpu.committedInsts <<< sim_insts) *? ** > > > Regards, > > Ali Chaker > _______________________________________________ > gem5-users mailing list > [email protected] > http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users > > > > _______________________________________________ > gem5-users mailing list > [email protected] > http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users >
_______________________________________________ gem5-users mailing list [email protected] http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users
