By using the "timing" cpu, you are effectively using something that is like
an idealized 1-wide, in-order cpu model.  So the maximum possible IPC would
be one and with cache accesses, etc it should be expected to be much lower
than one.

For relative comparisons, especially papers not looking explicitly at core
architecture, it's probably ok.  Be aware that compared to an out of order
cpu model, in-order cpu models have no way of overlapping/hiding memory
latency with other useful work (like in the shadow of a miss to an L2
cache).  So the relative performance differences vs an out of order cpu
model may be skewed.


On Thu, Feb 28, 2013 at 5:46 PM, Rodrigo Reynolds Ramírez <
[email protected]> wrote:

>  Hello everyone,
>
> I am working with cache replacement policies and I want to compare the IPC
> between these algorithms, I am using a --cpu-type=timing. For comparing
> the performance I am using the IPC, I am calculating it as 
> system.cpu.committedInsts/system.cpu.numCycles,
> but the values I got are different from papers and the ones I got from
> other simulator.
>
> The timing model is valid for calculating the IPC?
> I have read in the forum some people talking about low IPC for h264,
> actually all my spec2006 benchmarks have a low IPC.
>
> Thanks in advance
>                                  Rodrigo
>
> _______________________________________________
> 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

Reply via email to