If you do look at the patch, you'll see that it's a pretty minor hook
in the BaseSimpleCPU code, and doesn't touch atomic.* or timing.* at
all.  Plus warming up state like caches, bpred, etc. in a fast
functional mode is a key aspect of SMARTS; you spend more time in
warmup than in detailed simulation, so you'd lose a huge chunk of the
perf advantage if you did warmup in O3.

Steve

On Sat, Jul 10, 2010 at 12:53 AM, Gabe Black <[email protected]> wrote:
> I haven't looked at the code yet, but I'm wary of unsimplifying the
> simple CPU. It's been headed that way anyway, but just as much as
> functionally correct execution required. I'd like to see us resimplify
> it somehow instead. I'm not familiar with SMARTS, but would you be able
> to have some short lead in/warm up period with O3 before you started
> taking stats? That sounds very familiar so maybe you already do?
>
> Gabe
>
> Timothy Jones wrote:
>> -----------------------------------------------------------
>> This is an automatically generated e-mail. To reply, visit:
>> http://reviews.m5sim.org/r/48/
>> -----------------------------------------------------------
>>
>> Review request for Default.
>>
>>
>> Summary
>> -------
>>
>> SimpleCPU: Allow Simple CPUs to warm a branch predictor by creating a pointer
>> to a branch predictor in their class and a warm function within the branch
>> predictor itself.  The Simple CPU branch predictor is also provided in the
>> parameters so that it can be exposed to python.
>>
>>
>> Diffs
>> -----
>>
>>   src/cpu/pred/base.hh PRE-CREATION
>>   src/cpu/pred/bpred_unit.hh PRE-CREATION
>>   src/cpu/pred/bpred_unit_impl.hh PRE-CREATION
>>   src/cpu/simple/BaseSimpleCPU.py 249f174e6f37
>>   src/cpu/simple/base.hh 249f174e6f37
>>   src/cpu/simple/base.cc 249f174e6f37
>>
>> Diff: http://reviews.m5sim.org/r/48/diff
>>
>>
>> Testing
>> -------
>>
>>
>> Thanks,
>>
>> Timothy
>>
>> _______________________________________________
>> m5-dev mailing list
>> [email protected]
>> http://m5sim.org/mailman/listinfo/m5-dev
>>
>
> _______________________________________________
> m5-dev mailing list
> [email protected]
> http://m5sim.org/mailman/listinfo/m5-dev
>
_______________________________________________
m5-dev mailing list
[email protected]
http://m5sim.org/mailman/listinfo/m5-dev

Reply via email to