|
Message
From: Matjaz Breskvar<phoenix@o...>
Date: Thu Feb 17 02:55:39 CET 2005
Subject: [openrisc] or1ksim stable_0_1_0
* György 'nog' Jeney (nog@s...) wrote:
> > Just to make sure, are you running this on otherwise idle machine ?
>
> Yep, the machine was otherwise idle.
>
> > Do you have any daemon that renices processes hard on cpu or similar ?
>
> Nope, nothing of this sort.
>
> Just to be absolutely sure that nothing is running, I rebooted into runlevel 1
> and ran the bechmarks again. The times are more eaven, but --enable-fsim still
> shows a minor slowdown. I still don't understand why that is...
>
> +-----+-------------------------+----------------------------+
> | Run | Real with --enable-fsim | Real without --enable-fsim |
> +-----+-------------------------+----------------------------+
> | 1 | 2m42.967s | 2m42.081s |
> | 2 | 2m42.965s | 2m42.026s |
> +-----+-------------------------+----------------------------+
>
> Again, the raw data appears below:
>
> Without --enable-fsim:
> real 2m42.081s
> user 2m33.980s
> sys 0m7.980s
>
> real 2m42.026s
> user 2m34.860s
> sys 0m7.170s
>
> With --enable-fsim:
> real 2m42.967s
> user 2m35.380s
> sys 0m7.580s
>
> real 2m42.965s
> user 2m35.540s
> sys 0m7.420s
My testruns also showed similar results (few secundes faster execution at
~3 min total running time without --enable-fsim).
So go ahead throw it out, i guess nobody uses it anyway. (if you do now is
the time to shout and convince us why we should keep it).
best regards,
p.
PS. Here is one example run...
with --enable-fsim
0xc004c978, npc: 0xc004c97c, cycles 2639603090, #108829136
exit(-1042605588)
@reset : cycles 0, insn #0
@exit : cycles 2639617720, insn #108829667
diff : cycles 2639617720, insn #108829667
real 3m49.526s
user 2m51.581s
sys 0m5.193s
without --enable-fsim
0xc004c978, npc: 0xc004c97c, cycles 2639603090, #108829136
exit(-1042605588)
@reset : cycles 0, insn #0
@exit : cycles 2639617720, insn #108829667
diff : cycles 2639617720, insn #108829667
real 3m47.118s
user 2m51.156s
sys 0m5.121s
|
 |
Browse
