[Intel-gfx] [PATCH i-g-t] benchmarks/wsim: Simulate and interpret .wsim

Tvrtko Ursulin tvrtko.ursulin at linux.intel.com
Tue Jul 10 16:05:18 UTC 2018


On 10/07/2018 16:40, Chris Wilson wrote:
> Quoting Tvrtko Ursulin (2018-07-10 16:38:14)
>>
>> On 10/07/2018 14:47, Chris Wilson wrote:
>>> Quoting Tvrtko Ursulin (2018-05-11 09:31:52)
>>>>
>>>> On 11/05/2018 08:11, Chris Wilson wrote:
>>>>> A little tool I've been meaning to write for a while... Convert the
>>>>> .wsim into their dag and find the longest chains and evaluate them on an
>>>>> simulated machine.
>>>>
>>>> Very cool!
>>>
>>> Would you care to ack it in its current form, knowing that we will fix
>>> it whenever we find a corner case interesting enough to study?
>>
>> I think just a little bit more polish is needed. I've just tried it out
>> and polish I can see is:
>>
>> 1. -h / --help with just a brief comment on what the tool does and usage.
>>
>> 2. Lead stats displayed for each cmdline argument with parsed filename.
>>
>> 3. Replace result section names "Single client", "Simulated clients" and
>> item names "total/ideal/packed" with a more descriptive text which would
>> hopefully be somewhat self-explanatory what the numbers represent.
>>
>> 4. Report garbage input (and unknown wsim commands) instead of reporting
>> some numbers for instance for ./sim_wsim README :)
>>
>> 5. #ifdef 0 graphviz ? add --graphviz cmddline option?
>>
>> 6. Oh.. and meson makefiles.. :)
> 
> But if it was upstream, you would just contribute those improvements
> yourself.
> 
> So you are saying that this is just not useful?

No, just that I think we have to have some minimum standard before we 
can make it upstream. If you want I can implement these bits and send a 
v3. Then it will just be a matter of finding a third person to review it 
since we will both be authors. :)

Regards,

Tvrtko


More information about the Intel-gfx mailing list