For 2 and 3 below, what are the  other stats changes? I'd expect some small
difference due to introduction of a slightly new feature in the inorder
model and I think it'd be OK to commit in those sparc cases. I'm more
worried if you are breaking for alpha or mips since those are more tested
in the inorder model.

For 1), like Andreas has said, I've had situations where I've had to go
debug on zizzer and it turns out that the issue is some unitialized
variable that only manifests itself when run on zizzer. Since it's a branch
predictor chagne, I'd theorize it's also probably safe to commit given the
suspicions of myself + Andreas on that workload.

With that said, are there any other obstacles to committing this change?

I push the issue because I dont get as much spare time to go and debug gem5
like I use to, so I'm hoping the effort I put into finding that small
difference in the inorder branch predictor makes it into an actual
changeset.


>
> Seems like I will not be able to keep my promise. Following differences
> were found after running all the regression tests --
>
> 1. tests/long/se/20.parser/ref/arm/linux/o3-timing/
>   sim_insts                       505237723  505237723          0    +0.00%
>   sim_ops                         569624283  569624283          0    +0.00%
>   sim_ticks                      199845137000 199893963500   48826500
> +0.02%
>
> 2. tests/quick/se/00.hello/ref/sparc/linux/inorder-timing/
> None to sim_insts, sim_ops, sim_ticks, but other stats change.
>
> 3. tests/quick/se/02.insttest/ref/sparc/linux/inorder-timing/
> None to sim_insts, sim_ops, sim_ticks, but other stats change.
>
> Ali, is this fine with you?
>





>
> --
> Nilay
>
> _______________________________________________
> gem5-dev mailing list
> [email protected]
> http://m5sim.org/mailman/listinfo/gem5-dev
>



-- 
- Korey
_______________________________________________
gem5-dev mailing list
[email protected]
http://m5sim.org/mailman/listinfo/gem5-dev

Reply via email to